2.1 TAB 6 7 9 10 11
TBC 2 3 2 3 2 3 2 3 2 3
CAB 1200 1200 550 550 550
CBC 300 300 300 300 300 300 300 300 300 300
TAC 8 9 9 10 11 12 12 13 13 14
CAC 1500 1500 1500 1500 850 850 850 850 850 850
(a) Sample space of travel time from A to B = {6, 7, 9, 10, 11} Sample space of travel time from A to C = {8, 9, 10, 11, 12, 13, 14} (b) Sample space of travel cost from A to C = {850, 1500} (c) Sample space of TAC and CAC = {(8, 1500), (9, 1500), (10, 1500), (11, 850), (12, 850), (13, 850), (14, 850)}
2.2 (a) Since the possible values of settlement for Pier 1 overlap partially with those of Pier 2, it is possible that both Piers will have the same settlement. Hence, the minimum differential settlement is zero. The maximum differential settlement will happen when the settlement of Pier 2 is 10 cm and that of Pier 1 is 2 cm, which yields a differential settlement of 8 cm. Hence, the sample space of the differential settlement is zero to 8 cm. (b) If the differential settlement is assumed to be equally likely between 0 and 8 cm, the probability that it will be between 3 and 5 cm is equal to
P=
5−3 2 = = 0.25 8−0 8
2.3 (a)
(b) Wind direction
E 90o
E
E
60o
E
30o Wind Speed 0
15
35
45
(c)
A and B are not mutually exclusive A and C are not mutually exclusive
2.4
(a) Inflow 6’ 6’ 6’ 7’ 7’ 7’ 8’ 8’ 8’
Possible water level Outflow Inflow – Outflow + 7’ 5’ 8’ 6’ 7’ 7’ 6’ 5’ 9’ 6’ 8’ 7’ 7’ 5’ 10’ 6’ 9’ 7’ 8’
Hence possible combinations of inflow and outflow are (6’,5’), (6’,6’), (6’,7’), (7’,5’), (7’,6’), (7’,7’), (8’,5’), (8’,6’) and (8’,7’). (b) The possible water levels in the tank are 6’, 7’, 8’, 9’ and 10’. (c) Let E = at least 9 ft of water remains in the tank at the end of day. (8’, 5’) and (8’, 6’) are favourable to the event E. So P(E) =
3 1 = . 9 3
Sample points (7’, 5’),
2.5 (a) Locations Locations Load at B Load at C MA Probability of W2 of W1 – – 0 0 0 0.15x0.2=0.03 – B 500 0 5,000 0.045 – C 0 500 10,000 0.075 B – 200 0 2,000 0.05 B B 700 0 7,000 0.075 B C 200 500 12,000 0.125 C – 0 200 4,000 0.12 C B 500 200 9,000 0.18 C C 0 700 14,000 0.30
E1
X X X X X
E2
E3
X X X X
X X
X X
(b) E1 and E2 are not mutually exclusive because these two events can occur together, for example, when the weight W2 is applied at C, MA is 10,000 ft-lb; hence both E1 and E2 will occur. (c) The probability of each possible value of MA is tabulated in the last column of the table above. (d) P(E1) = P(MA > 5,000) = 0.075 + 0.075 + 0.125 + 0.18 + 0.30 = 0.755 P(E2) = P(1,000 ≤ MA ≤ 12,000) = 0.045 + 0.075 + 0.05 + 0.075 + 0.12 + 0.18 = 0.545 P(E3) = 0.05 + 0.075 = 0.125 P(E1 ∩ E2) = P(5,000 < MA ≤ 12,000) = 0.075 + 0.075 + 0.18 = 0.33 P(E1 ∪ E2) = 1 – 0.03 = 0.97 P( E 2 ) = 0.03 + 0.125 + 0.3 = 0.455 P( P ( E2 ) = 1 − P ( E ) = 1 − 0.545 = 0.455
2.6 (a) Let A1 = Lane 1 in Route A requires major surfacing A2 = Lane 2 in Route A requires major surfacing B1 = Lane 1 in Route B requires major surfacing B2 = Lane 2 in Route B requires major surfacing P(A1) = P(A2) = 0.05 P(B1) = P(B2) = 0.15
P(A2⏐A1) = 0.15 P(B2⏐B1) = 0.45
P(Route A will require major surfacing) = P(A) = P(A1∪A2) = P(A1) + P(A2) - P(A2⏐A1) P(A1) = 0.05 + 0.05 - (0.15)(0.05) = 0.0925 P(Route B will require major surfacing) = P(B) = P(B1∪B2) = P(B1) + P(B2) - P(B2⏐B1) P(B1) = 0.15 + 0.15 - (0.45)(0.15) = 0.2325 (b) P(route between cities 1 and 3 will require major resurfacing) = P(A∪B) = P(A) + P(B) - P(A)P(B) = 0.0925 + 0.2325 – (0.0925)(0.2325) = 0.302
2.7 P(Di) = 0.1 Assume condition between welds are statistically independent (a) P( D1 D 2 D3 ) = P( D1 )P( D 2 )P( D3 ) = 0.9 x 0.9 x 0.9 = 0.729 (b) P(Exactly two of the three welds are defective) = P( D1 D2D3 ∪ D1 D 2 D3 ∪ D1 D2 D3 ) = P( D1 D2D3) + P(D1 D 2 D3) + P(D1 D2 D3 ) since the three events are mutually exclusive. Hence, the probability become P = 0.9x0.1x0.1 + 0.1x0.9x0.1 + 0.1x0.1x0.9 = 0.027 (c) P(all 3 welds defective) = P(D1 D2 D3) = P3(D) = (0.1)3 = 0.001
2.8 P(E1) = 0.8; P(E2) = 0.7; P(E3) = 0.95 P(E3⏐ E 2 ) = 0.6; assume E2 and E3 are statistically independent of E1 (a) A = (E2 ∪ E3) E1 B = ( E 2 ∪ E 3 ) E1
or E 2 E 3 ∪ E1
(b) P(B) = P( E1 ∪ E 2 E 3 ) = P( E1 ) + P( E 2 E 3 ) – P( E1 E 2 E 3 ) = 0.2 + P( E 3 ⏐ E 2 )P( E 2 ) – P( E1 )P( E 2 E 3 ) = 0.2 + 0.4x0.3 – 0.2x(0.4x0.3) = 0.296 (c) P(casting ⏐concrete production not feasible at site) = P((E2 ∪ E3) E1⏐ E 2 )
=
P( E 2 ∪ E 3 ) E1 E 2 P( E 2 )
= 0.8 x 0.6 =0.48
=
P( E1 E 2 E 3 ) P( E 2 )
=
P( E1 ) P( E 3 E 2 ) P( E 2 ) P( E 2 )
2.9 E1, E2, E3 denote events tractor no. 1, 2, 3 are in good condition respectively (a) A = only tractor no. 1 is in good condition = E1 E 2 E 3 B = exactly one tractor is in good condition = E1 E 2 E 3 ∪ E1 E2 E 3 ∪ E1 E 2 E3 C = at least one tractor is in good condition = E1 ∪ E2 ∪ E3 (b) Given P(E1) = P(E2) = P(E3) = 0.6 P( E 2 ⏐ E1 ) = 0.6 or P( E 2 ⏐ E 3 ) = 0.6 P( E 3 ⏐ E1 E 2 ) = 0.8 or P( E1 ⏐ E 2 E 3 ) = 0.8 P(A) = P(E1 E 2 E 3 ) = P(E1⏐ E 2 E 3 ) P( E 2 E 3 ) = [1 – P( E1 ⏐ E 2 E 3 )] P( E 2 ⏐ E 3 ) P( E 3 ) = (1 - 0.8)(0.6)(0.4) = 0.048 Since E1 E 2 E 3 , E1 E2 E 3 and E1 E 2 E3 are mutually exclusive; also the probability of each of these three events is the same, P(B) = 3 x P(E1 E 2 E 3 ) = 3 x 0.048 = 0.144 P(C) = P(E1 ∪ E2 ∪ E3) = 1 – P( E1 ∪ E 2 ∪ E 3 ) = 1 – P( E1 E 2 E 3 ) = 1 – P( E1 ⏐ E 2 E 3 ) P( E 2 ⏐ E 3 ) P( E 3 ) = 1 – 0.8x0.6x0.4 = 0.808
2.10 (a) The event “both subcontractors will be available” = AB, hence since P(A∪B) = P(A) + P(B) - P(AB) ⇒ P(AB) = P(A) + P(B) - P(A∪B) = 0.6 + 0.8 - 0.9 = 0.5 (b) P(B is available⏐A is not available) = P(B⏐ A ) =
P( BA ) P( A )
while it is clear from the following Venn diagram that P( BA ) = P(B) - P(AB).
A
AB
B
A Hence
P( BA ) P( B) − P( AB) = 1 − P( A) P( A ) = (0.8 - 0.5)/(1 - 0.6) = 0.3/0.4 = 0.75 (c) (i) If A and B are s.i., we must have P(B⏐A) = P(B) = 0.8. However, using Bayes’ rule, P(B⏐A) = P(AB)/P(A) = 0.5/0.6 = 0.8333 So A and B are not s.i. (A’s being available boosts the chances that B will be available) (ii) From (a), P(AB) is nonzero, hence AB ≠ ∅, i.e. A and B are not m.e. (iii) Given: P(A∪B) = 0.9 ⇒ A∪B does not generate the whole sample space (otherwise the probability would be 1), i.e. A and B are not collectively exhaustive.
2.11 (a) Let L, SA, SB denote the respective events “leakage at site”, “seam of sand from X to A”, “seam of sand from X to B”. Given probabilities: P(L) = 0.01, P(SA) = 0.02, P(SB) = 0.03, P(SB⏐SA) = 0.2, Also given: independence between leakage and seams of sand, i.e. P(SB⏐L) = P(SB), P(L⏐SB) = P(L), P(SA⏐L) = P(SA), P(L ⏐SA) = P(L) The event “water in town A will be contaminated” = L∩SA, whose probability is P(L∩SA) = P(L⏐ SA) P(SA) = P(L)P(SA) = 0.01×0.02 = 0.0002. (b) The desired event is (L SA) ∪ (L SB), whose probability is P(L SA) + P(L SB) - P(L SA L SB) = P(L)P(SA) + P(L)P(SB) - P(L SA SB) = P(L) [P(SA) + P(SB) - P(SA SB)] = P(L) [P(SA) + P(SB) - P(SB⏐SA) P(SA)] = 0.01 (0.02 + 0.03 - 0.2×0.02) = 0.00046
2.12 Let A,B,C denote the respective events that the named towns are flooded. Given probabilities: P(A) = 0.2, P(B) = 0.3, P(C) = 0.1, P(B | C) = 0.6, P(A | BC) = 0.8, P( AB | C ) = 0.9, where an overbar
denotes compliment of an event.
(a) P(disaster year) = P(ABC) = P(A | BC)P(BC) = P(A | BC)P(B | C)P(C) = 0.8×0.6×0.1 = 0.048 (b) P(C | B) = P(BC) / P(B) = P(B | C)P(C) / P(B) = 0.6×0.1÷0.3 = 0.2 (c) The event of interest is A∪B∪C. Since this is the union of many items, we can work with its compliment instead, allowing us to apply De Morgan’s rule and rewrite as P(A∪B∪C) = 1 – P( A ∪ B ∪ C ) = 1 – P( ABC ) = 1 – P( AB | C )P( C ) = 1 – 0.9×(1 – 0.1) = 1 – 0.81 = 0.19
by De Morgan’s rule,
2.13 [(L∪M) G]] ∪[(LM) G ]
(a)
C=
(b)
Since (L∪M)G is contained in G, it is mutually exclusive to (LM) G which is contained in G . Hence P(C) is simply the sum of two terms, P(C) =
P[(L∪M) G]] + P[(L∩M) G ]
= P(LG∪MG) + P(L)P(M G ) = P(LG) + P(MG) – P(LMG) + P(L)P(M| G )P( G ) = P(L)P(G) + P(M|G)P(G) – P(L)P(M|G)P(G)+ P(L)P(M| G )P( G ) = 0.7×0.6 + 1×0.6 – 0.7×1×0.6 + 0.7×0.5×0.4 = 0.74
(c)
P( L |C) = P( L C)/P(C) = P( L {[(L∪M) G]} ∪[(LM) G ])) / P(C) = P[ L (L∪M) G]] / P(C) impossible event = P( L MG) / P(C) = P( L )P(MG) / P(C) = P( L )P(M|G)P(G) / P(C) = 0.3×1×0.6 / 0.74 ≅ 0.243
since ( L L)M G is an
2.14 (a) Note that the question assumes an accident either occurs or does not occur at a given crossing each year, i.e. no more than one accident per year. There were a total of 30 + 20 + 60 + 20 = 130 accidents in 10 years, hence the yearly average is 130 / 10 = 13 accidents among 1000 crossings, hence the yearly probability of accident occurring at a given crossing is 13 = 0.013 (probability per year) 1000
(b) Examining the data across the “Day” row, we see that the relative likelihood of R compared to S is 30:60, hence P(S | D) = 60/90 = 2/3 (c) Let F denote “fatal accident”. We have 50% of (30 + 20) = 0.5×50 = 25 fatal “run into train” accidents, and 80% of (60 + 20) = 0.8×80 = 64 fatal “struck by train” accidents, hence the total is P(F) = (25 + 64) / 130 ≅ 0.685 (d) (i) (ii)
D and R are not mutually exclusive; they can occur together (there were 30 run-into-train accidents happened in daytime); If D and R are, we must have P(R | D) = P(R), but here P(R | D) = 30 / 90 = 1/3, while P(R) = (30 + 20) / 130 = 5/13, so D and R are not statistically independent.
2.15 F = fuel cell technology successfully marketable S = solar power technology successfully marketable F and S are statistically independent Given P(F) = 0.7; P(S) = 0.85 (i)
P(energy supplied) = P(F∪S) = P(F) + P(S) – P(F)P(S) = 0.7 + 0.85 – 0.7x0.85 = 0.955
(ii)
P(only one source of energy available) = P(F S ∪ F S) = P(F S ) + P( F S) = (0.7)(1-0.85) + (1-0.7)(0.85) = 0.36
2.16 a. E1 = Monday is a rainy day E2 = Tuesday is a rainy day E3 = Wednesday is a rainy day Given P(E1) = P(E2) = P(E3) = 0.3 P(E2⏐ E1) = P(E3⏐ E2) = 0.5 P(E3⏐ E1E2) = 0.2 b.
P(E1E2) = P(E2⏐ E1) P(E1) = 0.5x0.3 = 0.15
c.
P(E1E2 E 3 ) = P( E 3 ⏐E1E2) P(E2⏐ E1) P(E1) = (1-0.2)(0.5)(0.3) = 0.12
d.
P(at least one rainy day) = P(E1 ∪ E2 ∪ E3) = P(E1) + P(E2) + P(E3) - P(E1E2) - P(E2E3) - P(E1E3) + P(E1E2E3) = 0.3 + 0.3 + 0.3 - 0.3x0.5 - 0.3x0.5 – 0.1125 + 0.3x0.5x0.2 =0.52 Where P(E1E3)=P(E3|E1)P(E1)=0.375x0.3=0.1125 P(E3|E1)= P(E3|E2E1)P(E2|E1)+ P(E3| E 2 E1)P( E 2 |E1)
=0.15x0.15+0.3x0.5 =0.375
2.17 Let A, B, C denote events parking lot A, B and C are available on a week day morning respectively Given: P(A) = 0.2; P(B) = 0.15; P(C) = 0.8 P(B⏐ A ) = 0.5;
P(C⏐ A B ) = 0.4
(a)
P(no free parking) = P( A B ) = P( A ) x P( B ⏐ A ) = 0.8 x 0.5 = 0.4
(b)
P(able to park)
= 1 – P(not able to park) = 1 – P( A B C ) = 1 – P( A B )xP( C ⏐ A B ) = 1 – 0.4x(1-0.4) = 1 – 0.24 =0.76
(c)
P(free parking⏐able to park) P(A∪B⏐A∪B∪C) =
P[( A ∪ B )( A ∪ B ∪ C )] P( A ∪ B ∪ C )
=
P( A ∪ B) P( A ∪ B ∪ C )
=
1 − P( AB) 0.76
=
1 − 0.4 = 0.789 0.76
2.18 C = Collapse of superstructure E = Excessive settlement Given: P(E) = 0.1; P(C) = 0.05 P(C⏐E) = 0.2 (a)
P(Failure) = P(C ∪ E) = P(C) + P(E) – P(C⏐E)P(E) = 0.05 + 0.1 – 0.2x0.1 = 0.13
(b)
P( EC⏐E ∪ C)
=
P(( EC ) ∩ ( E ∪ C )) P( E ∪ C )
= P(EC) / 0.13 = 0.2x0.1 / 0.13 = 0.154
(c)
P(E C ∪ E C) = P(E C ) + P( E C) = P(E∪C) – P(EC) = 0.13 – 0.02 =0.11
2.19 M = failure of master cylinder W = failure of wheel cylinders B = failure of brake pads Given: P(M) = 0.02; P(W) = 0.05; P(B) = 0.5 P(MW) = 0.01 B is statistically independent of M or W
(a)
P(W M B ) = P(W M )P( B ) = [P(W) – P(MW)] [1 – P(B)] = (0.05 – 0.01)(1 – 0.5) = 0.02
(b)
P(system failure) = 1 – P(no component failure) = 1- P( M W B ) = 1- P( M W ) P( B ) = 1- {1 - P(M ∪ W)}P( B ) = 1- {1-[P(M)+P(W)-P(MW)]}P( B ) = {1 - [0.02+0.05-0.01]}(0.5) = 0.53
(c)
P(only one component failed) = P( W M B) + P( W M B ) + P(W M B ) P(W M B ) = 0.02 from part (a) also P( W M B ) = P( W M) x P( B ) = [P(M)-P(MW)]P( B ) = (0.02 – 0.01)x0.5 = 0.005
also P( W M B) = P( W M ) x P( B ) = [1-P(M)-P(W)+P(MW)]P(B)
= (1-0.02-0.005+0.01)x0.5 = 0.47
Since the three events W M B, W M B and W M B are all within the event of system failure P(only one component failure⏐system failure) = (0.47+0.005+0.02) / 0.53 = 0.934
2.20 E1 = Excessive snowfall in first winter E2 = Excessive snowfall in second winter E3 = Excessive snowfall in third winter (a)
P(E1) = P(E2) = P(E3) = 0.1 P(E2 ⏐E1) = 0.4 = P(E3 ⏐E2) P(E3 ⏐E1 E2) = 0.2
(b)
P(E1 ∪ E2) = P(E1)+P(E2)-P(E1 E2) = 0.1 + 0.1 – 0.4x0.1 = 0.16
(c)
P(E1E2E3) = P(E1) P(E2 ⏐E1) P(E3 ⏐E1 E2) = 0.1x0.4x0.2 =0.008
(d)
P( E2 ⏐ E1 ) = ? Since P( E1 ∪ E2 ) = P( E1 )+P( E2 )-P( E1 E2 ) = P( E1 )+P( E2 )-P( E2 ⏐ E1 )P( E1 ) = 0.9+0.9-P( E2 ⏐ E1 ) 0.9 and from given relationship, P( E1 ∪ E2 ) = 1- P(E1E2) = 1- P(E1) P(E2 ⏐E1) = 1-0.1x0.4 = 0.96 Therefore, 1.8-0.9 P( E2 ⏐ E1 ) = 0.96 and P( E2 ⏐ E1 ) = 0.933
2.21 H1, H2, H3 denote first, second and third summer is hot respectively Given: P(H1) = P(H2) = P(H3) = 0.2 P(H2 ⏐H1) = P(H3 ⏐E2) = 0.4 (a)
P(H1H2H3) = P(H1) P(H2 ⏐H1) P(H3 ⏐E2) = 0.2x0.4x0.4 = 0.032
(b)
P( H 2 ⏐ H 1 ) = ? Using the hint given in part (d) of P2.20, P( H 1 ) + P( H 2 ) - P( H 2 ⏐ H 1 )P( H 1 ) = P( H 1 ∪ H 2 ) = 1 – P(H1H2) = 1 - P(H2 ⏐H1) P(H1) we have, 0.8 + 0.8 - P( H 2 ⏐ H 1 )(0.8) = 1 – 0.4x0.2 therefore, P( H 2 ⏐ H 1 ) = 0.85
(c)
P(at least 1 hot summer) = 1 – P(no hot summers) = 1 – P( H 1 H 2 H 3 ) = 1 – P( H 1 )P( H 2 ⏐ H 1 ) P( H 3 ⏐ H 2 ) = 1 – 0.8x0.85x0.85 = 0.422
2.22 A = Shut down of Plant A B = Shut down of Plant B C = Shut down of Plant C Given: P(A) = 0.05, P(B) = 0.05, P(C) = 0.1 P(B⏐A) = 0.5 = P(A⏐B) C is statistically independent of A and B (a)
P(complete backout⏐A) = P(BC⏐A) = P(B⏐A)P(C) = 0.5x0.1 = 0.05
(b)
P(no power) = P(ABC) = P(AB)P(C) = P(B⏐A)P(A)P(C) = 0.5x0.05x0.1 =0.00025
(c)
P(less than or equal to 100 MW capacity) = P (at most two plants operating) = 1 – P(all plants operating) = 1 – P( A B C ) = 1 – P( C )P( A B ) = 1 – P( C ) [1-{P(A)+P(B)- P(B⏐A)P(A)}] = 1 – 0.9[1-{0.05+0.05-0.5x0.05}] = 0.1675
2.23 P(Damage) = P(D) = 0.02 Assume damages between earth quakes are statistically independent (a)
P(no damage in all three earthquakes) = P3(D) = 0.023 = 8x10-6
(b)
P( D1 D2) = P( D1 )P(D2) = 0.98x0.02 = 0.0196
2.24 (a)
Let A, B denote the event of the respective engineers spotting the error. Let E denote the event that the error is spotted, P(E) = P(A∪B) = P(A) + P(B) – P(AB) = P(A) + P(B) – P(A)P(B) = 0.8 + 0.9 – 0.8×0.9 = 0.98
(b)
“Spotted by A alone” implies that B failed to spot it, hence the required probability is P(AB’|E) = P(AB’∩E)/P(E) = P(E|AB’)P(AB’) / P(E) = 1×P(A)P(B’) / P(E) = 0.8×0.1 / 0.98 ≅ 0.082
(c)
With these 3 engineers checking it, the probability of not finding the error is P(C1’C2’C3’) = P(C1’)P(C2’)P(C3’) = (1 – 0.75)3 , hence P(error spotted) = 1 – (1 – 0.75)3 ≅ 0.984, which is higher than the 0.98 in (a), so the team of 3 is better.
(d)
The probability that the first error is detected (event D1) has been calculated in (a) to be 0.98. However, since statistical independence is given, detection of the second error (event D2) 2 still has the same probability. Hence P(D1 D2) = P(D1)P(D2) = 0.98 ≅ 0.960
2.25 L = Failure of lattice structure A = Failure of anchorage Given:
P(A) = 0.006 P(L⏐A) = 0.4 P(A⏐L) = 0.3 Hence, P(L) = P(L⏐A)P(A) / P(A⏐L) = 0.008
(a)
P(antenna disk damage) = P(A∪L) = P(A) + P(L) – P(A)P(L⏐A) = 0.4 + 0.008 - 0.006x0.4 = 0.406
(b)
P(only one of the two potential failure modes) = P(A L ) + P( A L) = P( L ⏐A)P(A) + P( A ⏐L)P(L)(0.008) = (1-0.4)(0.006) + (1-0.3)(0.008) = 0.0036 + 0.0056 =0.0092
(c)
P ( A D) =
P( A( A ∪ L)) P( A) 0.006 = = = 0.0148 P( A ∪ L) P ( A ∪ L) 0.406
2.26 P(F) = 0.01 P(A⏐ F ) = 0.1 P(A⏐F) = 1
(a)
FA, F A , F A, F A are set of mutually exclusive and collectively exhaustive events
(b)
P(FA) = P(A⏐F)P(F) = 1x0.01 = 0.01 P(F A ) = P( A ⏐F)P(F) = 0x0.01 = 0 P( F A) = P(A⏐ F )P( F ) = 0.1x0.99 = 0.099 P( F A ) = P( A ⏐ F )P( F ) = 0.9x0.99 = 0.891
(c)
P(A) = P(A⏐F)P(F) + P(A⏐ F )P( F ) = 1x0.01 + 0.1x0.99 = 0.109
(d)
P(F⏐A) =
P( A F ) P( F ) P( A)
=
0.01 = 0.0917 0.109
2.27 Given:
P(D) = 0.001 P(T⏐D) = 0.85; P(T⏐ D ) = 0.02
(a)
DT, D T , D T, D T
(b)
P(DT) = P(T⏐D)P(D) = 0.85x0.001 = 0.00085 P(D T ) = P( T ⏐D)P(D) = 0.15x0.001 = 0.00015 P( D T) = P(T⏐ D )P( D ) = 0.02x0.999 = 0.01998 P( D T ) = P( T ⏐ D )P( D ) = 0.98x0.999 = 0.979
(c)
P( D T ) =
P(T D) P( D) P(T D) P( D) + P(T D) P( D)
=
0.00085 = 0.0408 0.00085 + 0.01998
2.28 Given:
P(RA) = P(GA) = 0.5 P(RB) = P(GB) = 0.05 P(GB⏐GA) = 0.8 P(GLT) = 0.2 Signal at C is statistically independent of those at A or B.
(a)
E1 = RA∪ RB E2 = GLT E3 = RA GB ∪ GA RB
(b)
P(stopped at least once from M to Q) = 1 – P(no stopping at all from M to Q) = 1 – P(GA GB GLT) = 1 – P(GLT)P(GA)P(GB ⏐GA) = 1 – 0.2x0.5x0.8 = 0.92
(c)
P(stopped at most once from M to N) = 1 – P(stopped at both A and B) = 1 – P(RARB) = 1 - P(RA) P(RB⏐RA) From the hint given in P2.20, it can be shown that
P( RB R A ) =
1 1 [1 − P(G B G A ) P(G A )] − 1 = [1 − 0.8 × 0.5] − 1 = 0.2 P( R A ) 0 .5
Hence P(stopped at most once from M to N) = 1 – 0.5x0.2 = 0.9
2.29 (a)
P(WSC) = P(W)P(SC) = P(W)P(S|C)P(C) = 0.1×0.3×0.05 = 0.0015
(b)
P(W∪C) = P(W) + P(C) – P(WC) = P(W) + P(C) – P(W)P(C) = 0.1 + 0.05 – 0.1×0.05 = 0.145 P(W∪C| S ) = P[ S (W∪C)] / P( S ) = P( S W∪ S C) / P( S ) = [ P( S W) + P( S C) – P( S WC) ] / P( S ) = [ P( S )P(W) + P( S |C)P(C) – P(W)P( S C) ] / P( S ) = [0.8×0.1 + (1 – 0.3)×0.05 – 0.1×(1 – 0.3)×0.05] / (1 – 0.2) ≅ 0.139
(c)
(d)
P(nice winter day) = P( S W C) = P( W )P( S C) = P( W )P( S | C)P(C) = 0.9x0.7x0.05 = 0.0315
(e)
P(U) = P(U |CW)P(CW) + P(U | C W)P( C W) + P(U|C W ) P(C| W ) + P(U|| C W ) P( C W ) = 1xP(W)P(C) + 0.5xP(W)P( C ) + 0.5xP(C)P( W ) + 0xP( C )P( W ) = 1x0.1x0.05 + 0.5x0.1x0.95 + 0.5x0.05x0.9 = 0.075
2.30 (a)
Let L and B denote the respective events of lead and bacteria contamination, and C denote water contamination. P(C) = P(L∪B) = P(L) + P(B) – P(LB) = P(L) + P(B) – P(L)P(B)∵ L and B are independent events = 0.04 + 0.02 – 0.04×0.02 = 0.0592
(b)
P(L B | C) = P(CL B ) / P(C) = P(C | L B )P(L B ) / P(C), but P(C | L B ) = 1 (∵ lead alone will contaminate for sure) and P(L B ) = P(L)P( B ) (∵ statistical independence), hence the probability = 1×P(L)P( B ) / P(C) = 1×0.04×(1 – 0.02) / 0.0592 ≅ 0.662
2.31 C 0.1 0.1 0.1 0.2 0.2 0.2
Water level 0 10 20 0 10 20
W (x103 lb) 100 210 320 100 210 320
F (x103 lb) 10 21 32 20 42 64
(a)
Sample space of F = {10, 20, 21, 32, 42, 64}
(b)
Let H be the horizontal force in 103 lb P(sliding) = P(H>F) = P(F<15) = 0.1
(c)
P(sliding) = P(F<15) P(H=15) + P(F<20) P(H=20) = 0.1x0.5 + 0.1x0.1 = 0.06
Probability 0.5x0.2=0.1 0.5x0.4=0.2 0.5x0.4=0.2 0.5x0.2=0.1 0.5x0.4=0.2 0.5x0.4=0.2
2.32 Given:
(a)
P(W) = 0.9 P(H) = 0.3 P(E) = 0.2 P(W⏐H) = 0.6 E is statistically independent of W or H
I = E( W ∪H) II = E W H
(b)
P[E( W ∪H)] = P(E) P( W ∪H) = P(E) x [P( W )+P(H)-P( W ⏐H)P(H)] = 0.2x(0.1+0.3-0.4x0.3) = 0.056
P( E W H)
= P( E ) P( W ⏐H) P(H) = 0.8x0.4x0.3 = 0.096
(c1)
Since P(W⏐H) = 0.6 ≠ 0, W and H are not mutually exclusive. Since P(W⏐H) = 0.6 ≠ 0.9 = P(W), W and H are not statistically independent
(c2)
I II = E( W ∪H) ( E W H) = (E E )( W ∪H)( W H) Since E E is an empty set, I and II are mutually exclusive.
From the above Venn Diagram, the union of I and II does not make up the entire sample space. Hence, I and II are not collectively exhaustive.
(d)
Since I and II are mutually exclusive, P(leakage) = P(I) + P(II) = 0.056 + 0.096 = 0.152
2.33 Given:
P(E) = 0.15 P(G) = 0.1 P(O) = 0.2 P(E⏐O) = 2x0.15 = 0.3 G is statistically independent of E or O
(a)
P(shortage of all three sources) = P(EGO) = P(G) P(E⏐O) P(O) = 0.1x0.3x0.2 = 0.006
(b)
P(G ∪ E) = 1-P( G E ) = 1- P( G ) P( E ) = 1-0.9x0.85 = 0.235
(c)
P(GO⏐E) = P(GOE)/P(E) = 0.006/0.15 = 0.04
(d)
P(at least 2 sources will be in short supply) = P(EG∪GO∪EO) = P(EG)+P(GO)+P(EO)-P(EGO)-P(EGO)-P(EGO)+P(EGO) = P(G)P(E)+P(G)P(O)+P(E⏐O)P(O)-2P(EGO) = 0.1x0.15 + 0.1x0.2 + 0.3x0.2 - 2x0.006 = 0.083
2.34 A, B, C denote failure of component A, B, C respectively N and H denote normal and ultra-high altitude respectively Given: P(A⏐N) = 0.05, P(B⏐N) = 0.03, P(C⏐N) = 0.02 P(A⏐H) = 0.07, P(B⏐H) = 0.08, P(C⏐H) = 0.03 P(N) = 0.6, P(H) = 0.4 P(B⏐A) = 2 x P(B) C is statistically independent of A or B P(system failure) = P(S) = P(S⏐N)P(N) + P(S⏐H)P(H) P(S⏐N) = P(A∪B∪C⏐N) = P(A⏐N) + P(B⏐N) + P(C⏐N) - P(AB⏐N) - P(BC⏐N) - P(AC⏐N) + P(ABC⏐N) = 0.05 + 0.03 + 0.02 – 2x0.03x0.03 – 0.03x0.02 – 0.05x0.02 + 0.02x2x0.03x0.03 = 0.096636 Similarly, P(S⏐H) = P(A⏐H) + P(B⏐H) + P(C⏐H) - P(AB⏐H) - P(BC⏐H) - P(AC⏐H) + P(ABC⏐H) = 0.07 + 0.08 + 0.03 – 2x0.08x0.08 – 0.08x0.03 – 0.07x0.03 + 0.03x2x0.08x0.08 = 0.16308 Hence P(system failure) = 0.96636x0.6 + 0.16308x0.4 ≅ 0.123
P( B S ) = =
P[ B( A ∪ B ∪ C )] P( B ) P ( B N ) P( N ) + P( B H ) P ( H ) = = P( S ) P( S ) P( S )
0.03 × 0.6 + 0.08 × 0.4 = 0.172 0.123
2.35 Given:
P(C) = 0.5 P(W) = 0.3 P(W⏐C) = 0.4 U=C∪W
(a)
Since P(W⏐C) = 0.4 ≠ 0.3=P(W) W and C are not statistically independent
(b)
P(U) = P(C ∪ W) = P(C) + P(W) – P(W⏐C) P(C) = 0.5 + 0.3 - 0.4x0.5 = 0.6
(c)
P(C W ) = P( W ⏐C) P(C) = (1-0.4)x0.5 = 0.3
(d)
P (CW C ∪ W ) =
P (W C ) P(C ) 0.4 × 0.5 P[CW (C ∪ W )] P(CW ) = = = = 0.333 0.6 0.6 P(C ∪ W ) P (C ∪ W )
2.36 A, B, C denote route A, B, C are congested respectively Given: P(A) = 0.6, P(B) = 0.6, P(C) = 0.4 P(A⏐B) = P(B⏐A) = 0.85 C is statistically independent of A or B P(L⏐ABC) = 0.9 P(L⏐ ABC ) = 0.3
(a)
P(A B C )+ P( A B C )+ P( A B C) = P( B ⏐A)P(A)P( C )+P( A ⏐B)P(B)P( C )+P( A B )P(C)
but P( A B ) = 1-P(A ∪B) = 1-P(A)-P(B)+ P(B⏐A)P(A) = 1 - 0.6 - 0.6 + 0.85x0.6 = 0.31 Hence, P(exactly one route congested) = 0.15x0.6x0.6 + 0.15x0.6x0.6 +0.31x0.4 = 0.232
(b)
P(Late) = P(L⏐ABC)P(ABC)+ P(L⏐ ABC ) P( ABC ) But P(ABC) = P(AB)P(C) = P(B⏐A)P(A)P(C) = 0.85x0.6x0.4 = 0.204 Hence, P(L) = 0.9x0.204 + 0.3x(1-0.204) = 0.422
(c)
If C is congested, P(ABC) = 0.85x0.6x1 = 0.51 P(L) = 0.9x0.51 + 0.3x0.49 = 0.606
2.37 (a) Let subscripts 1 and 2 denote “after first earthquake” and “after second earthquake”. Note that (i) occurrence of H1 makes H2 a certain event; (ii) H, L and N are mutually exclusive events and their union gives the whole sample space. P(heavy damage after two quakes) = P(H2 H1) + P(H2 L1) + P(H2 N1) = P(H2 | H1) P(H1) + P(H2 | L1)P(L1)+ P(H2 | N1)P(N1) = 1×0.05 + 0.5×0.2 + 0.05×(1 – 0.2 – 0.05) ≅ 0.188 (b) The required probability is [P(H2 L1) + P(H2 N1)] / P(heavy damage after two quakes) = (0.5×0.2 + 0.05×0.75) / (1×0.05 + 0.5×0.2 + 0.05×0.75) ≅ 0.733 (c) In this case, one always starts from an undamaged state, the probability of not getting heavy damage at any stage is simply (1 – 0.05) = 0.95 (not influenced by previous condition). Hence P(any heavy damage after 3 quakes) = 1 – P(no heavy damage in each of 3 quakes) 3 = 1 – 0.95 ≅ 0.143. Alternatively, one could explicitly sum the probabilities of the three mutually exclusive events, P(H) = P(H1) + P(H2 H1’) + P(H3 H2’ H1’) = 0.05 + 0.05×0.95 + 0.05×0.95×0.95 ≅ 0.143
2.38 Given:
P(L) = 0.6, P(A) = 0.3, P(H) = 0.1 S denote supply adequate, i.e. no shortage P(S⏐L) = 1, P(S⏐A) = 0.9, P(S⏐H) = 0.5
(a)
P(S) = P(S⏐L)P(L)+P(S⏐A)P(A)+P(S⏐H)P(H) = 1x0.6 + 0.9x0.3 + 0.5x0.1 = 0.92
(b)
P( A S ) =
(c)
P( S A) P( A) P(S )
=
0.1 × 0.3 = 0.375 1 − 0.92
P(shortage in at least one or of the next two months) = 1 – P(no shortage in next two months) = 1 – 0.92x0.92 = 0.154
2.39 (a)
P(shipment accepted on a given day) = P(at most 1 defective panel) = 0.2 + 0.5 = 0.7
(b)
P(exactly one shipment will be rejected in 5 days) = 5xP(0.7)4(0.3) = 0.36
(c)
P(acceptance of shipment on a given day) = P(A) = P(A⏐D=0)P(D=0)+ P(A⏐D=1)P(D=1)+ P(A⏐D=2)P(D=2) in which D = number of defective panels on a given day P(A⏐D=0) = 1 P(A⏐D=1) = 1 P(A⏐D=2) = 1-P(both defective panels detected) = 1 – 0.8x0.8 = 0.36 Hence, P(A) = 0.2 + 0.5 + 0.36x0.3 = 0.808
2.40 The possible scenario of the working condition and respective probabilities are as follows:
(a)
P(completion on schedule) = P(E) = P(E⏐G) P(G) + P(E⏐N) P(N) + P(E⏐BC) P(BC) + P(E⏐B C ) P(B C ) = 1x0.2 + 0.9x0.4 + 0.8x0.4x0.5 + 0.2x0.4x0.5 = 0.76
(b)
P(Normal weather⏐completion) = P(N⏐E) = P(E⏐N) P(N) / P(E) = 0.9x0.4 / 0.76 = 0.474
2.41 Let L, N, H denote the event of low, normal and high demand respectively; also O and G denote oil and gas supply is low respectively. (a)
Given normal energy demand, probability of energy crisis = P(E⏐N) = P(O∪G) = P(O)+P(G)-P(OG) = P(O)+P(G)-P(G⏐O)P(O) = 0.2 + 0.4 – 0.5x0.2 = 0.5
(b)
P(E) = P(E⏐L) P(L) + P(E⏐N) P(N) + P(E⏐H) P(H) = P(OG)x0.3 + 0.5x0.6 +1x0.1 = P(G⏐O)P(O)x0.3 + 0.3 + 0.1 = 0.43
2.42 Given:
P(H=1) = 0.2, P(H=2) = 0.05, P(H=0) = 0.75 Where H = number of hurricanes in a year P(J=H=1) = P(D⏐H=1) = 0.99 P(J=H=2) = P(D⏐H=2) = 0.80 Where J and D denote survival of jacket and deck substructure respectively Assume J and D are statistically independent
(a)
For the case of one hurricane, i.e. H=1 P(damage) = 1-P(JD⏐H=1) = 1- P(J⏐H=1) P(D⏐H=1) = 1 - 0.99x0.99 = 0.0199
(b)
For next year where the number of hurricanes is not known. P(no damage) = P(JD⏐H=0)P(H=0)+ P(JD⏐H=1)P(H=1)+ P(JD⏐H=2)P(H=2) = 1x0.75 + 0.99x0.99x0.2 + 0.8x0.8x0.05 = 0.75 + 0.196 + 0.032 = 0.978
(c)
P(H=0⏐JD) =
P( JD H = 0) P( H = 0) 1 × 0.75 = = 0.767 P ( JH ) 0.978
2.43 (a) Sample space = {FA, HA, EA, FB, HB, EB} in which FA, FB denote fully loaded truck in lane A, B respectively HA, HB denote half loaded truck in lane A, B respectively EA, EB denote empty truck in lane A, B respectively (b)
Assume the events of F, H and E are equally likely P(full⏐truck in lane B) = 1/3
(c)
Let C denote event of critical stress P(C) = P(C⏐ FA)P(FA)+ P(C⏐ HA)P(HA)+ P(C⏐ EA)P(EA) +P(C⏐ FB)P(FB)+ P(C⏐ HB)P(HB)+ P(C⏐ EB)P(EB) But P(C⏐ FA)=1, P(C⏐ HA)=0.4, P(C⏐ EA)=0 P(C⏐ FB)=0.6, P(C⏐ HB)=0.1, P(C⏐ EB)=0
1 5 5 × = 3 6 18 1 P( E B ) = 16
P(FA) = P(fully loaded⏐truck in lane A) P(truck in lane A) =
1 1 1 1 × = , P( H B ) = , 3 6 18 16 5 1 P( H A ) = , P( E A ) = 18 18
Similarly P(FB) =
Hence, P(C) = 1x5/18 + 0.4x5/18 + 0 + 0.6x1/18 + 0.1x1/18 + 0 =0.428
2.44 S denotes shear failure B denotes bending failure D denotes diagonal cracks Given: P(D⏐S) = 0.8 P(D⏐B) = 0.1 5% of failure are in shear mode 95% of failure are in bending mode (a)
P(D) = P(D⏐S)P(S) + P(D⏐B)P(B) = 0.8x0.05 + 0.1x0.95 = 0.135
(b)
P( S D) =
P( D S ) P( S ) P( D)
=
0.8 × 0.05 = 0.296 0.135
Since the failure in the shear mode was only 29.6%, it does not exceed the threshold value of 75% required, immediate repair is not recommended.
2.45 (a) Let A, D, I denote the respective events that a driver encountering the amber light will accelerate, decelerate, or be indecisive. Let R denote the event that s/he will run the red light. The given probabilities are P(A) = 0.10, P(D) = 0.85, P(I) = 0.05, also the conditional probabilities P(R | A) = 0.05, P(R | D) = 0, P(R | I) = 0.02 By the theorem of total probability, P(R) = P(R | A)P(A) + P(R | D)P(D) + P(R | I)P(I) = 0.05×0.10 + 0 + 0.02×0.05 = 0.005 + 0.001 = 0.006 (b) The desired probability is P(A | R), which can be found by Bayes’ Theorem as P(A | R) = P(R | A)P(A) / P(R) = 0.05×0.10 / 0.006 = 0.005 / 0.006 ≅ 0.833 (c) Let V denote “there exists vehicle waiting on the other street”, where P(V) = 0.60 and P(V’) = 0.40; and let C denote “Cautious driver in the other vehicle”, where P(C) = 0.80 and P(C’) = 0.20. The probability of collision is P(collision) = P(collision | V)P(V) + P(collision | V’)P(V’) But the second term is zero (since there is no other vehicle to collide with), while P(collision | V) in the first term is P(collision | C)P(C) + P(collision | C’)P(C’) = (1 – 0.95)×0.80 + (1 – 0.80)×0.20 = 0.05×0.80 + 0.20×0.20 = 0.08, hence P(collision) = P(collision | V)P(V) = 0.08×0.60 = 0.048 (d) 100000 vehicles × 5% = 5000 vehicles are expected to encounter the yellow light annually. Out of these 5000 vehicles, 0.6% (i.e. 0.006) are expected to run a red light, i.e. 5000×0.006 = 30 vehicles. These 30 dangerous vehicles have 0.048 chance of getting into a collision (i.e.
accident), hence 30×0.048 = 1.44 accidents caused by dangerous vehicles can be expected at the intersection per year
2.46 A, B, C denote branch A, B, C will be profitable respectively E denotes bonus received P(A) = 0.7, P(B) = 0.7, P(C) = 0.6 P(A⏐B) = 0.9 = P(B⏐A) C is statistically independent of A or B (a)
P(Exactly two branches profitable) = P(AB C )+P(A B C)+P( A BC) = P(A⏐B)P(B)P( C )+P( B ⏐A)P(A)P(C)+P( A ⏐B)P(B)P(C) = 0.9x0.7x0.4 + 0.1x0.7x0.6 + 0.1x0.7x0.6 = 0.336
(b)
P(at least two branches profitable) = P(T) = P(ABC) + 0.336 = P(A⏐B)P(B)P(C) + 0.336 = 0.9x0.7x0.6 + 0.336 = 0.714 P(bonus received) = P(E) = P(E⏐T)P(T)+P(E⏐ T )P( T ) = 0.8x0.714 + 0.2x0.286 = 0.628
(c)
Given A is not profitable, P(T⏐ A ) = P(BC⏐ A ) = P(B⏐ A )P(C) But P ( B A) =
P( A B) P( B) P( A)
=
0.1 × 0.7 = 0.233 0.3
Hence P(T⏐ A ) = 0.233x0.6 = 0.1398 P(bonus received) = 0.8x0.1398 + 0.2x0.8602 = 0.284
2.47 P(D) = P(difficult foundation problem) = 2/3 P(F) = P(project in Ford County) = 1/3 = 0.333 P(I) = P(project in Iroquois County) = 2/5 =0.4 P(D⏐I) = 1.0 P(D⏐F) = 0.5 P(C) = P(project in Champaign County) = 1 – 0.333 – 0.4 = 0.267
(a)
P(F D ) = P( D ⏐F)P(F) = 0.5x0.333 = 0.167
(b)
P(C D ) = P( D ⏐C)P(C)
But P( D ) = P( D ⏐C)P(C)+ P( D ⏐F)P(F)+ P( D ⏐I)P(I) 0.333 = P( D ⏐C)x0.267 + 0.5x0.333 + 0x0.4
∴ P( D ⏐C)x0.267 = 0.167 → P( D ⏐C) = 0.625
Hence P(C D ) = 0.625x0.167 = 0.104
(c)
P( I D) =
P( D I ) P( I ) P( D)
=
0 × 0.4 = 0.0 0.104
2.48 P(E) = 0.001, P(L) = 0.002, P(T) = 0.0015 P(L⏐E) = 0.1 T is statistically independent of E or L (a)
P(ELT) = P(L⏐E)P(E)P(T) = 0.1x0.001x0.0015 = 1.5x10-6
(b)
P(E∪L∪T) = 1-P( E L T ) = 1-P( T )P( E ∪ L ) = 1 – 0.9985x[1-P(E)-P(L)+P(L⏐E)P(E)] = 1 – 0.0015[1-0.001-0.002+0.1x0.001] = 1 – 0.9985x0.9971 = 0.0044
(c)
P( E E ∪ L ∪ T ) =
P ( E ( E ∪ L ∪ T )) P( E ) 0.001 = = = 0.227 P( E ∪ L ∪ T ) 0.0044 0.0044
2.49 (a) Let E1, E2 denote the respective events of using 100 and 200 units, and S denote shortage of material. P(S) = P(S | E1)P(E1) + P(S | E2)P(E2) = 0.1×0.6 + 0.3×0.4 = 0.06 + 0.12 = 0.18 (b) Using Bayes’ theorem with the result from part (a), P(E1 | S) = P(S | E1)P(E1) / P(S) = 0.06 / 0.18 = 1/3
2.50 Let H and S denote Hard and Soft ground, respectively, and let L denote a successful landing. Given probabilities: P(L | H) = 0.9; P(L | S) = 0.5; P(H) = 3P(S) , but since ground is either hard or soft ⇒ P(H) = 0.75, P(S) = 0.25 (a) Using theorem of total probability, P(L) = P(L | H)P(H) + P(L | S)P(S) = 0.9×0.75 + 0.5×0.25 = 0.8 (b) Let E denote “penetration”. Given: P(E | S) = 0.9; P(E | H) = 0.2 (i) The “updated” probability of hard ground, P’(H) ≡ P(H | E) = P(E | H)P(H) / [P(E | H)P(H) + P(E | S)P(S)] by Bayes’ theorem = 0.2×0.75 / (0.2×0.75 + 0.9×0.25) = 0.4 (ii) Using the updated probabilities P’(H) = 0.4, P’(S) = 1 - 0.4 = 0.6, the updated probability of a successful landing now becomes P’(L) = P(L | H)P’(H) + P(L | S)P’(S) = 0.9×0.4 + 0.5×0.6 = 0.66
2.51 P(C) = 0.1, P(S) = 0.05 Where C, S denote shortage of cement and steel bars respectively P(S⏐ C ) = 0.5x0.05 = 0.025
(a)
P(S∪C) = P(S)+P(C)- P(C⏐S)P(S) But P( C ⏐S) =
P( S C ) P(C ) P(S )
=
0.025 × 0.9 = 0.45 0.05
Hence P(S∪C) = 0.05 + 0.1 – 0.55x0.05 = 0.1225
(b)
P(C S ∪ C S) = P(C S )+P( C S) = P(C∪S) – P(CS) from Venn diagram = 0.1225 - P(C⏐S)P(S) = 0.1225 – 0.55x0.05 = 0.095
(c)
P(S S ∪ C ) =
P( S ( S ∪ C )) P(S ) 0.05 = = = 0.408 P(S ∪ C ) P( S ∪ C ) 0.1225
2.52 Let A and B be water supply from source A and B are below normal respectively P(A) = 0.3, P(B) = 0.15 P(B⏐A) = 0.3 P(S⏐A B ) = 0.2, P(S⏐ A B) = 0.25, P(S⏐ A B ) = 0,
P(S⏐AB) = 0.8
Where S denotes event of water shortage (i)
P(A∪B) = P(A)+P(B)- P(B⏐A)P(A) = 0.3 + 0.15 - 0.3x0.3 = 0.36
(ii) P(A B ∪ A B) = P(A∪B) – P(AB) = P(A)+P(B)-2P(B⏐A)P(A) = 0.3 + 0.15 –2x0.3x0.3 =0.22
(iii) P(S) = P(S⏐A B )P(A B )+P(S⏐ A B)P( A B)+P(S⏐ A B )P( A B )+P(S⏐AB)P(AB) But P ( A B ) =
P( B A) P( A) P( B)
=
0.3 × 0.3 = 0.6 0.15
Hence P(S) = 0.2x0.7x0.3 + 0.25x0.4x0.15 + 0 + 0.8x0.3x0.3 = 0.129
(iv) P ( AB S ) =
(v)
P( S AB) P( AB) P( S )
=
0.8 × 0.3 × 0.3 = 0.558 0.129
P( A ⏐ S ) = P( A B⏐ S )+P( A B ⏐ S )
P( S AB ) P( AB) =
P( S )
+
P( S AB) P( AB) P( S )
But P( A B) = [1-P(A⏐B)] P(B) = 0.4x0.15 = 0.06
P( A B ) = 1-P(AB)-P( A B)-P(A B ) = 1 – 0.3x0.3 – 0.06 – 0.7x0.3
Hence P ( A S ) =
0.75 × 0.06 1 × 0.64 + = 0.786 0.871 0.871
2.53 Let W be weather favorable F be field work completed on schedule C be computation completed T1 be computer 1 available T2 be computer 2 available
Given: P(F⏐W) = 0.9,
P(F⏐ W ) = 0.5
P( W ) = 0.6 P(T1) = P(T2) = 0.7 P(C⏐T1 T2 ) = P(C⏐ T1 T2) = 0.6 P(C⏐T1 T2) = 0.9,
(a)
P(C⏐ T1 T2 ) = 0.4
P(F) = P(F⏐W)P(W) + P(F⏐ W )P( W ) = 0.9x0.4 + 0.5x0.6 = 0.66
(b)
P(C) = P(C⏐T1 T2) P(T1 T2)+ P(C⏐T1 T2 ) P(T1 T2 )+ P(C⏐ T1 T2) P( T1 T2)+ P(C⏐ T1 T2 ) P( T1 T2 ) = 0.9x0.7x0.7 + 0.6x0.7x0.3 + 0.6x0.3x0.7 + 0.4x0.3x0.3 = 0.729
(c)
P(Deadline met) = P(FC) = P(F)P(C) = 0.66x0.729 = 0.481
2.54 (a)
T = wait time in queue (in min.) N = number of trucks in queue P(T<5⏐N=2) = P(loading time for both trucks ahead will take only 2 min. each) = 0.5x0.5 = 0.25
(b)
P(T<5) = P(T<5⏐N=0)P(N=0)+ P(T<5⏐N=1)P(N=1)+ P(T<5⏐N=2)P(N=2) Note that if the number of trucks in queue is 3 or more, the waiting time will definitely exceed 5. Hence those items will not contribute any probabilities. Hence P(T<5) = 1x0.175 + 1x0.125 + 0.25x0.3 = 0.375
2.55 Let O1 and O2 denote the events of truck 1 (in one lane) and 2 (in the other lane) being overloaded, respectively, and let D denote the event of bridge damage. The given probabilities are P(O1) = P(O2) = 0.1; P(D | O1O2) = 0.3; P(D | O1O2’) = P(D | O1’O2) = 0.05; P(D | O1’O2’) = 0.001. (a) P(D) can be determined by the theorem of total probability as P(D| O1O2)P(O1O2) + P(D| O1O2’)P(O1O2’) + P(D| O1’O2)P(O1’O2) + P(D| O1’O2’)P(O1’O2’) = 0.3×0.1×0.1 + 0.05×0.1×(1 – 0.1) + 0.05×(1 – 0.1)×0.1 + 0.001×(1 – 0.1)×(1 – 0.1) ≅ 0.0128 (b) P(O1 ∪ O2 | D) = 1 – P[ (O1 ∪ O2 )’ | D] (but De Morgan’s rule says (O1 ∪ O2 )’ = O1’O2’) = 1 – P(O1’O2’| D) = 1 – P(D | O1’O2’)P(O1’O2’) / P(D) = 1 – 0.001×0.9×0.9 / 0.01281 ≅ 0.937 (c) The first alternative reduces all those conditional probabilities in (a) by half, hence the P(D) also reduces by half to 0.0128 / 2 = 0.0064. If one adopts the second alternative, P(D) becomes P(D| O1O2)P(O1O2) + P(D| O1O2’)P(O1O2’) + P(D| O1’O2)P(O1’O2) + P(D| O1’O2’)P(O1’O2’) = 0.3×0.06×0.06 + 0.05×0.06×(1 – 0.06) + 0.05×(1 – 0.06)×0.06 + 0.001×(1 – 0.06)×(1 – 0.06) ≅ 0.0076 Hence one should take the first alternative (strengthening the bridge) to minimize P(D).
2.56 (a) Let A = “presence of anomaly” and D = “detection of anomaly by geophysical techniques”. We are given P(D | A) = 0.5 ⇒ P(D’ | A) = 1 – 0.5 = 0.5, and also P(D | A’) = 0 ⇒ P(D’| A’) = 1. We need P(A | D’) = P(D’ | A)P(A) / P(D’) by Bayes’ theorem, in which P(D’) = P(D’ | A)P(A) + P(D’ | A’)P(A’) = 0.5×0.3 + 1×0.7 = 0.85, hence P(A | D’) = 0.5×0.3 / 0.85 ≅ 0.176 (b) (i) With the updated anomaly probability P*(A) = 0.15/0.85 (thus P*(A’) = 0.7/0.85), and also a better detection probability P(D | A) = 0.8 (thus P(D’ | A) = 0.2), the probability of having no anomaly, given no detection, is now P**(A’) = P(A’ | D’) = P(D’ | A’)P*(A’) / P(D’) = 1×(0.7/0.85) / [1×(0.7/0.85) + 0.2×(0.15/0.85)] ≅ 0.959 (ii) Total probability of a safe foundation = P(safe | A)P**(A) + P(safe | A’)P**(A’) = 0.80×(1 – 0.959) + 0.9999×0.959 ≅ 0.992 (iii)
A failure probability of p means that out of a large number (N) of similar systems, p×N of them are expected to fail. The total failure cost would be p×N×$1000000, which when divided into N gives the “average” or expected cost per system = p×$1000000. Hence we have the formula Expected loss = (probability of failure) × (failure loss) = (1 – 0.992)×$1000000 ≅ $8300 If the system is known for sure to be anomaly free, however, it has only 1 – 0.9999 = 0.0001 chance of failure, in which case the expected loss would be 0.0001×$1000000 = $100. Hence, ($8300 - $100) = $8200 in expected loss is saved when the site can be verified to be anomaly free
2.57 Given:
P(G) = 0.3, P(A) = 0.6, P(B) = 0.1 P(F⏐G) = 0.001, P (F⏐A) = 0.01,
(a)
P(F) = P(F⏐G)P(G)+ P(F⏐A)P(A)+ P(F⏐B)P(B) = 0.001x0.3 + 0.01x0.6 + 0.1x0.1 = 0.0003 + 0.06 +0.01 = 0.0703
(b)
T = passing the test P(T⏐G) = 0.9, P(T⏐A) = 0.7,P(T⏐B) = 0.2
(i)
P(G⏐T) =
=
(ii)
P(F⏐B) = 0.1
P(T⏐G ) P(G ) P(T⏐G ) P(G ) + P(T⏐A) P( A) + P (T⏐B) P( B)
0.9 × 0.3 0.27 = = 0.38 0.9 × 0.3 + 0.7 × 0.6 + 0.2 × 0.1 0.71
P’(F) = P(F⏐G)P’(G)+ P(F⏐A)P’(A)+ P(F⏐B)P’(B) But P’(G) = P(G⏐T) = 0.38 P’(A) = P(A⏐T) = 0.7x0.6 / 0.71 = 0.59 P’(B) = P(B⏐T) = 0.2x0.1 / 0.71 = 0.03 Hence P’(F) = 0.001x0.38 + 0.01x0.59 + 0.1x0.03 = 0.00928
2.58 Given:
P(L) = 15/(15+4+1) = 15/20 = 0.75 P(M) = 4/20 = 0.20 P(H) = 1/20 = 0.05 P(P) = 0.2 where P denotes poorly constructed P(D⏐PL) = 0.1, P(D⏐WL) = 0,
P (D⏐PM) = 0.5, P(D⏐PH) = 0.9 P (D⏐WM) = 0.05, P(D⏐WH) = 0.2
(a)
P(D⏐W) = P(D⏐WM)P(L)+P(D⏐WM)P(M)+ P(D⏐WH)P(H) = 0x0.75 + 0.05x0.2 + 0.2x0.05 = 0.02
(b)
P(D) = P(D⏐W)P(W)+ P(D⏐P)P(P) But P(D⏐P) = P(D⏐PL)P(L)+P(D⏐PM)P(M)+ P(D⏐PH)P(H) = 0.1x0.75 + 0.5x0.2 + 0.9x0.05 = 0.22 Hence P(D) = 0.02x0.8 + 0.22x0.2 = 0.06 In other words, 6% of the buildings will be damaged.
(c)
P ( P⏐D) =
P( D⏐P) P( P) 0.22 × 0.2 = = 0.733 P( D) 0.06
2.59 (a) Note that any passenger must get off somewhere, so the sum over any row is one. Let Oi denote “Origin at station i” and Di denote “Departure at station i”, where i = 1,2,3,4. P(D3) = P(D3 | O1)P(O1) + P(D3 | O2)P(O2) + P(D3 | O2)P(O2) = 0.3×0.25 + 0.3×0.15 + 0.1×0.25 = 0.145 (b) For a passenger boarding at Station 1, his/her trip length would be 4, 9 or 14 minutes for departures at stations 2, 3 or 4, respectively. These lengths have respective probabilities 0.1, 0.3 and 0.6, hence the expected trip length is (4×0.1 + 9×0.3 + 14×0.6) miles = 11.5 miles (c) Let E = “trip exceeds 10 miles”, P(E) = P(E | O1)P(O1) + P(E | O2)P(O2) + P(E | O3)P(O3) + P(E | O4)P(O4) = P(D4 | O1) P(O1) + P(D1 | O2) P(O2) + P(D1 ∪ D2| O3) P(O3) + P(D2 ∪ D3| O4) P(O4) = 0.6×0.25 + 0.6×0.15 + (0.5 + 0.1)×0.35 + (0.1 + 0.1)×0.25 = 0.5 (d) P(O1 | D3) = P(D3 | O1) P(O1) / P(D3) = 0.3×0.25 / 0.145 ≅ 0.517
2.60 P(A) = 0.6, P(B) = 0.3, P(C) = 0.1 P(D⏐A) = 0.02, P (D⏐B) = 0.03,
(a)
P(A⏐D) =
=
(b)
P(D⏐C) = 0.04
P( D⏐A) P( A) P( D⏐A) P ( A) + P( D⏐B) P( B ) + P( D⏐C ) P(C )
0.02 × 0.6 = 0.48 0.02 × 0.6 + 0.03 × 0.3 + 0.04 × 0.1
P(A∪B⏐D) = P(A⏐D)+P(B⏐D) Since P(B⏐D) = 0.03x0.3 / 0.025 = 0.36 P(A∪B⏐D) = 0.48+0.36 = 0.84
2.61 P(winging project A) = P(A) = 0.5 P(B) = 0.3 P(A⏐B) = 0.5x0.5 = 0.25 P(B⏐A) = 0.5x0.3 = 0.15 (a)
P(A∪B) = P(A)+P(B)-P(A⏐B)P(B) = 0.5 + 0.3 – 0.25x0.3 = 0.725
(b)
P( AB A ∪ B) =
(c)
P ( A AB ∪ AB) =
P( B A) P( A) 0.85 × 0.5 P( AB( A ∪ B )) P( AB) = = = = 0.586 P( A ∪ B) P( A ∪ B) 0.725 0.725
P( A( AB ∪ AB)) P( A B ∪ AB)
=
(d)
=
P( AB) P( AB) + P( AB)
0.85 × 0.5 =0.654 0.725 − 0.15 × 0.5
P(C⏐E) = 0.75, P(C⏐ E ) = 0.5 Where C denotes completion of project A on time P(C) = P(C⏐E)P(E)+P(C⏐ E )P( E ) = 0.75x0.5 + 0.5x0.5 =0.625
(e)
P ( E⏐C ) =
P (C⏐E ) P ( E ) 0.75 × 0.5 = = 0.6 P(C ) 0.625
=
P( B A) P( A) P( A ∪ B) − P( AB)
2.62 Given:
P(A) = P(poor aggregate) = 0.2 P(W⏐A) = P(poor workmanship⏐poor aggregate) = 0.3 P(A⏐W) = 015
(a)
P(W) = P(A)P(W⏐A) / P(A⏐W) = 0.2x0.3 / 0.15 = 0.4
(b)
P(A∪W) = P(A)+P(W)- P(W⏐A)P(A) = 0.2 + 0.4 - 0.3x0.2 = 0.54
(c)
P(A W ∪ A W) = P(A∪W)-P(AW) = 0.54 - 0.3x0.2 = 0.48
(d)
Given P(defective concrete A W ) = P(D ⏐A W ) = 0.15 P(D⏐ A W) = 0.2,
P(D⏐AW) = 0.8,
P(D⏐ A W ) = 0.05
P(D) = P(D ⏐A W ) P(A W )+ P(D⏐ A W) P( A W)+ P(D⏐AW) P(AW)+ P(D⏐ A W ) P( A W )
But
P(A W ) = 0.7x0.2 = 0.14,
P( A W) = 0.85x0.4 = 0.34
P(AW) = 0.3x0.2 = 0.06,
P( A W ) = 1-0.14-0.34-0.06 = 0.46
Hence P(D) = 0.15x0.14 + 0.2x0.34 + 0.8x0.06 + 0.05x0.46 = 0.192
(e)
P ( AW⏐D) =
P( D⏐AW ) P ( AW ) 0.048 = = 0.25 P( D) 0.192
2.63
(a) Given probabilities: P(A | S) = 0.2, P(A | L) = 0.6, P(A | N) = 0.05. Since S, L, N are mu exclusive and collectively exhaustive, P(N) = 0.7 and P(S) = 2P(L) lead to P(S) = 0.2, P(L) = Hence P(A) = P(A | L)P(L) + P(A | S)P(S) + P(A | N)P(N) = 0.6×0.1 + 0.2×0.2 + 0.05×0.7 = 0.135
(b) Let E denote “weak material Encountered by boring”; we are given P(E | L) = 0.8, P(E | S) = 0.3 | N) = 0. (i) The total probability of NOT encountering any weak material, P(E’) = P(E’ | L)P(L) + P(E’ | S)P(S) + P(E’ | N)P(N) = (1 – 0 .8)×0.1 + (1 – 0.3)×0.2 + 1×0.7 = 0.86, Hence, applying Bayes’ theorem, P(L | E’) = P(E’ | L)P(L) / P(E’) = [1 – P(E | L)]P(L) / [1 – P(E)] = (1 – 0.8)×0.1 / 0.86 ≅ 0.023 This is P*(L), the updated probability of a large weak zone in the light of new informatio (ii)
(iii)
Similar to (i), P(S | E’) = P(E’ | S)P(S) / P(E’) = (1 – 0.3)×0.2 / 0.86 ≅ 0.163 This is P*(S), the updated probability of a small weak zone. The “updated” probability of A is calculated using P*(S) and P*(L), P*(A) = P(A | L)P*(L) + P(A | S)P*(S) + P(A | N)P*(N) Note that P(A | L), P(A | N), P(A | S) all remains their original values as the settlement-soil condition dependence is not affected by the borings, hence P*(A) = 0.6×0.0233 + 0.2×0.163 + 0.05×(1 – 0.0233 – 0.163) ≅ 0.087
2.64 Let A, B denote occurrence of earthquake in the respective cities, and D denote damage to dam. (a) P(A∪B) = P(A) + P(B) – P(AB)
= P(A) + P(B) – P(A)P(B) since A and B are independent events = 0.01 + 0.02 – 0.01×0.02 = 0.0298 (b)
Given: P(D|A B ) = 0.3, P(D| A B) = 0.1, P(D|AB) = 0.5, want: P(D) Theorem of total probability gives P(D) = P(D|A B )P(A B ) + P(D| A B)P( A B) + P(D|AB)P(AB) (note: P(D| A B ) = 0 so it is not included) = P(D|A B )P(A)P( B ) + P(D| A B)P( A )P(B) + P(D|AB)P(A)P(B) = 0.3×0.01×0.98 + 0.1×0.99×0.02 + 0.5×0.01×0.02 = 0.00502
(c)
Now that B has dropped out of the picture, P(D) = P(D|A)P(A) + P(D| A )P( A ) = 0.4×0.01 + 0 = 0.004 < 0.00502, the new site is preferred since the yearly probability of incurring damages is about 20% lower.
(d)
Let Di denote “damage due to event i”, where i = E(arthquake), L(andslide), S(oil being poor). P(DE∪DL∪DS) = 1 – P[ D E ∪ D L ∪ DS ]
= 1 – P( D E D L DS )
by De
Morgan’s rule, = 1 – P( D E )P( D L )P( DS )
one should move
independence, = 1 – (1 – 0.004) (1 – 0.002)(1 – 0.001) ≅ 0.006986 > 0.00502, the site in this case, as the new site has a higher risk.
due to
2.65 A, B, C denote company A, B, C discover oil respectively P(A) = 0.4, P(B) = 0.6, P(C) = 0.2 P(A⏐B) = 1.2x0.4 = 0.48 C is statistically independent of B or A
(a)
P(A∪B∪C) = 1-P( A B C ) = 1-P( A B )P( C )
But P( A B ) = 1-P(A∪B) = 1-P(A)-P(B)+P(A⏐B)P(B) = 1 - 0.4 - 0.6 + 0.48x0.6 = 0.288 Hence P(A∪B∪C) = 1 – 0.288x0.8 = 0.77
P(C (A ∪ B ∪ C )) P(C ) = = 0.26 P(A ∪ B ∪ C ) 0.77
(b)
P (C⏐A ∪ B ∪ C ) =
(c)
P(A B C ∪ A B C ∪ A B C) = P(A B C )+P( A B C )+P( A B C)
But P ( B⏐A) =
P( A⏐B) P( B) 0.48 × 0.6 = = 0.72 0.4 P( A)
Hence, P(A B C ∪ A B C ∪ A B C) = 0.28x0.4x0.8 + 0.52x0.6x0.8 + 0.288x0.2 = 0.397
2.66 (a) Partition the sample space into AG, AB, PG, PB. We are given (i) P(D | AG) = 0, P(D | AB) = 0.5; (ii) P(D | PG) = 0.3, P(D | PB) = 0.9; (iii) P(A) = 0.3, P(P) = 0.7; (iv) P(B | A) = 0.2 ⇒ P(G | A) = 0.8; P(B | P) = 0.1 ⇒ P(G | P) = 0.9; (v) G and B are mutually exclusive and collectively exhaustive. Theorem of total probability gives P(D) = P(D|AG)P(AG) + P(D | AB)P(AB) + P(D | PG)P(PG) + P(D | PB | PB) = P(D|AG)P(G|A)P(A) + P(D|AB)P(B|A)P(A) + P(D| PG)P(G|P)P(P) + P(D|PB)P(B|P)P(P) = 0×0.8×0.3 + 0.5×0.2×0.3 + 0.3×0.9×0.7 + 0.9×0.1×0.7 = 0.282 (b) Bad weather can happen during AM or PM hours, hence we may rewrite B as the union of two mutually exclusive events, B = BA∪BP, thus P(B | D) = P(BA | D) + P(BP | D) = P(BAD) / P(D) + P(BPD) / P(D) = [ P(D | AB)P(B | A)P(A) + P(D | PB)P(B | P)P(P) ] / P(D) = (0.5×0.2×0.3 + 0.9×0.1×0.7) / 0.282 ≅ 0.330 (c) Of all morning flights, 20% will encounter bad weather. When they do, there is 50% chance of having a delay. Hence 20%×0.5 = 10% of morning flights will be delayed (note that good weather guarantees take-off without delay for morning flights). Or, in more formal notation, P(D | A) = P(DA) / P(A) = [P(D | AB)P(AB) + P(D | AG)P(AG)] / P(A) (where DA = D(A∪B) = DAB∪DAG was substituted) (but P(D | AG) = 0) = P(D | AB)P(AB) / P(A) = P(D | AB)P(B | A) = 0.5×0.2 = 0.1 (i.e. 10%)
2.67 P(M) = 0.05 where M = malfunction of machinery P(W) = 0.08 where W = carelessness of workers
P(D⏐M W ) = 0.1, P(D⏐ M W) = 0.2, P(D⏐MW) = 0.8 Assume M and W are statistically independent
(a)
P(D) = P(D ⏐M W ) P(M W )+ P(D⏐ M W) P( M W) +P(D⏐MW) P(MW)+ P(D⏐ M W ) P( M W ) = 0.1x0.05x0.92 + 0.2x0.95x0.08 + 0.8x0.05x0.08 + 0 = 0.023
(b)
P(W⏐D) = P(W M ⏐D) P(WM⏐D) = (0.0152+0.0032) / 0.023 = 0.8
2.68 Let A, B denote the events that these respective wells observed contaminants, and L denote that there was indeed leakage. Given: P (A | L) = 0.8, P (B | L) = 0.9, P (L) = 0.7, also these wells never false-alarm, i.e. P ( A | L ) = 1, P( B | L ) = 1. (a)
Given A , Bayes’ theorem yields P(L | A ) = P( A | L)P(L) / P( A ) where P( A ) = P( A | L)P(L) + P( A | L )P( L ) = 0.2×0.7 + 1×0.3 = 0.44, hence P(L | A ) = 0.2×0.7 / 0.44 ≅ 0.318
(b) (i)
(ii)
It will be convenient to first restrict our attention to within L, the only region in which A or B can happen. In here, PL(A) = 0.8 and PL(B) = 0.9, hence P (A∪B | L) = PL(A) + PL(B) – PL(A)PL(B) = 0.8 + 0.9 – 0.8×0.9 = 0.98, but this is only a relative probability with respect to L, hence the true probability P(A∪B) = P(A∪B | L)×P(L) = 0.98×0.7 = 0.686 A B is the event “both wells observed no contaminant”. Note that, even in L, there is a small piece of A’B’, with probability P( A B | L)P(L) = (1 – 0.8)×(1 – 0.9)×0.7 = 0.014, adding this to P( A B | L ) = 0.3 (since both wells never false alarm) gives the total probability P( A B ) = 0.014 + 0.3 = 0.314, hence P( L | A B ) = P( A B | L )P( L ) / P( A B ) = 1×0.3 / 0.314 ≅ 0.955
(a) In terms of the ability to confirm actual existence of leakage, P(B | L) = 0.9 > P(A | L) = 0.8, i.e. B is better. Also, in terms of the ability to infer safety, we have P(L | A ) ≅ 0.318 as calculated in part (a), while a similar calculation gives P(L | B ) = P( B | L)P(L) / P( B ) = 0.1×0.7 / (0.1×0.7 + 1×0.3) ≅ 0.189 Hence B gives a better assurance of non-leakage when it observes nothing. In short, B is better.
2.69 Let A and B denote having excessive amounts of the named pollutant, and H denote having a health problem due contaminated water. Let us partition the sample space into AB, A B, A B , A B . The total probability of having a health problem is P(H) = P(H | AB)P(AB) + P(H | A B)P( A B) + P(H | A B )P(A B ) + P(H | A B )P( A B ) where P(H | AB) = P(H | A B ) = 1 since existence of pollutant A causes health problem for sure; P(H | A B) = 0.2 since existence of pollutant B affects 20% of all people (even when A is absent) P(H | A B ) = 0 since there is no cause for any health problem; also P(AB) = P(B | A)P(A) = 0.5×0.1 = 0.05; P( A B) = P(B) – P(AB) = 0.2 – 0.05 = 0.15 (this is clear from a Venn diagram); P(A B ) = P( B | A)P(A) = (1 – 0.5)×0.1 = 0.05; P(H) = 1×0.05 + 0.2×0.15 + 1×0.05 = 0.13
2.70 (a)
P(winning at most one job) = P(winning zero job) + P(winning exactly one job) = P( H1H 2 H 3 B1B2 ) + P( H1H 2 H 3 B1B2 ) + P( H1H 2 H 3 B1B2 ) + P( H1H 2 H 3 B1B2 ) + P( H1H 2 H 3 B1B2 ) + P( H1H 2 H 3 B1B2 ) = (0.4)6 + (0.4)4(0.6)x5 = 0.0041 + 0.0768 = 0.081
(b)
P(win at least 2 jobs) = 1 – P(win at most one job) = 1 – 0.081 = 0.919
(c)
P(win exactly 1 highway job).P(win zero building jobs) = P[( H1H 2 H 3 )+P( H1H 2 H 3 )+P( H1H 2 H 3 )].P( B1B2 ) = (0.4)2 + (0.6)4x3x(0.4)2 = 0.0461
3.1 Total time T = A + B, which ranges from (3 + 4 = 7) to (5 + 6 = 11). Divide the sample space into A = 3, A = 4, and A = 5 (m.e. & c.e. events) ⇒ P(T = 7) =
∑
P(T = 7 | A = n)P(A = n)
n= 3, 4 ,5
=
∑
P(B = 7 - n)P(A = n)
n= 3, 4 ,5
= P(B =4)P(A = 3) = 0.2×0.3 = 0.06 Similarly P(T = 8) = P(B = 5)P(A = 3) + P(B = 4)P(A = 4) = 0.6×0.3 = 0.2×0.5 = 0.28 P(T = 9) = P(B = 6)P(A = 3) + P(B = 5)P(A = 4) + P(B = 4)P(A = 5) = 0.2×0.3 + 0.6×0.5 + 0.2×0.2 = 0.4 P(T = 10) = P(B = 6)P(A = 4) + P(B = 5)P(A = 5) = 0.2×0.5 + 0.6×0.2 = 0.22 P(T = 11) = P(B = 6)P(A = 5) = 0.2×0.2 = 0.04 Check: 0.06 + 0.28 + 0.4 + 0.22 + 0.04 = 1. The PMF is plotted as follows:
0.4 P(T = t)
0.28 0.22 0.06
7
0.04
8
9
10
11
t
3.2 Let X be the profit (in $1000) from the construction job. (a) P(lose money) = P(X < 0) = Area under the PDF where x is negative = 0.02×10 = 0.2 (b) Given event is X > 0 (i.e. money was made), hence the conditional probability, P(X > 40 | X > 0) = P(X > 40 ∩ X > 0) ÷ P(X > 0) = P(X > 40) ÷ P(X > 0) Let's first calculate P(X > 40): comparing similar triangles formed by the PDF and the x-axis (with vertical edges at x = 10 and at x = 40, respectively), we see that P(X > 40) = Area of smaller triangle 2 ⎛ 70 − 40 ⎞ =⎜ ⎟ × Area of larger triangle ⎝ 70 − 10 ⎠ = 0.52 [(70 - 10)×(0.02) ÷ 2] = 0.015 Hence the required probability P(X > 40 | X > 0) is 0.015 ÷ [10×0.02 + (70 - 10)×(0.02) ÷ 2] = 0.015 ÷ 0.08 = 0.1875
3.3 ∞
(a) Applying the normalization condition
∫f
X
( x)dx = 1
−∞
6
∫
⇒ c( x − 0
⇒ c=
x2 )dx = 1 ⇒ 6
6
⎡ x2 x3 ⎤ c⎢ − ⎥ = 1 ⎣ 2 18 ⎦ 0
18 = 1/6 9 × 36 − 6 3
(b) To avoid repeating integration, let’s work with the CDF of X, which is 1 ⎡ x2 x3 ⎤ FX(x) = ⎢ − ⎥ 6 ⎣ 2 18 ⎦ 2 3 = (9x – x )/108 (for x between 0 and 6 only) Since overflow already occurred, the given event is X > 4 (cms), hence the conditional probability P(X < 5 | X > 4) = P(X < 5 and X > 4) / P(X > 4) = P(4 < X < 5) / [1 – P(X ≤ 4)] = [FX(5) - FX(4)] / [1 - FX(4)] 2 3 2 3 2 3 = [(9×5 – 5 ) – (9×4 – 4 )]/ [108 – (9×4 – 4 )] = (100 – 80) / (108 – 80) = 20/28 = 5/7 ≅ 0.714 (c) Let C denote “completion of pipe replacement by the next storm”, where P(C) = 0.6. If C indeed occurs, overflow means X > 5, whereas if C did not occur then overflow would correspond to X > 4. Hence the total probability of overflow is (with ’ denoting compliment) P(overflow) = P(overflow | C)P(C) + P(overflow | C’)P(C’) = P(X > 5)×0.6 + P(X > 4)×(1 – 0.6) = [1 - FX(5)]×0.6 + [1 - FX(4)]×0.4 = (1 – 100/108)×0.6 + (1 – 80/108)×0.4 ≅ 0.148
3.4 (a) The median of X, xm, is obtained by solving the equation that defines xm, P(X ≤ xm) = 0.5 ⇒ FX(xm) = 0.5 ⇒ 1- (10 ÷ xm)4 = 0.5 ⇒ xm ≈11.9 (inches) (b) First, obtain the PDF of X: fX(x) = dFX/dx = 4×104x-5 for x ≥ 10, and zero elsewhere. Hence the expected amount of snowfall in a severe snow storm is ∞
E(X) =
∫ xf
X
( x )dx
−∞ ∞
=
∫ 4 × 10
4
x − 4 dx
10
= −
4 4 −3 ∞ 10 [ x ] 10 3
4 = × 10 4 ≈ 13.3 (inches) 3 (c) P(disastrous severe snowstorm) = P(X > 15) = 1 - P(X ≤ 15) = 1 - [1 - (10/15)4] = (2/3)4 ≈ 0.2 ∴About 20% of snow storms are disastrous. (d) Let N denote "No disastrous snow storm in a given year". Also, let E0, E1, E2 denote the respective events of experiencing 0,1,2 severe snow storms in a year. In each event, the (conditional) probability of N can be computed: P(N | E0) = 1 (if there's no severe snow storm to begin with, definitely there won't be any disastrous one); P(N | E1) = P(X ≤ 15) = 1 - (10/15)4 = 1 - (2/3)4; P(N | E2) = [P(X ≤ 15)]2 = [1 - (2/3)4 ]2 (statistical independence between storms); Noting that E0, E1 and E2 are m.e. and c.e., the total probability of N can be computed: P(N) = P(N | E0)P(E0) + P(N | E1)P(E1) + P(N | E2)P(E2) = 1×0.5 + [1 - (2/3)4]×0.4 + [1 - (2/3)4]2×0.1 ≈ 0.885
3.5 Let F be the final cost (a random variable), and C be the estimated cost (a constant), hence X=F/C is a random variable. (a) To satisfy the normalization condition, a
∫ 1
a
3 3 ⎡ − 3⎤ dx = ⎢ ⎥ = 3 − = 1 , 2 a x ⎣ x ⎦1
hence a = 3/2 =1.5 (b) The given event is “F exceeds C by more than 25%”, i.e. “F > 1.25C”, i.e. “F / C > 1.25”, whose probability is ∞
P(X > 1.25) =
∫f
X
( x) dx
1.25 1.5
1.5
3 ⎡ − 3⎤ dx = ⎢ ⎥ 2 x ⎣ x ⎦ 1.25 1.25 = -2 – (- 2.4) = 0.4 =
∫
(c) The mean, 1.5
E(X) =
3
∫x x
2
dx = [3 ln x ]1 ≅ 1.216, 1.5
1
while 1.5
2
E(X ) =
∫x 1
2
3 dx = 3(1.5 – 1) = 1.5, x2
with these, we can determine the variance 2 2 Var(X) = E(X ) – [E(X)] 2 = 1.5 – 1.216395324 = 0.020382415 ∴σX =
0.020382415 ≅ 0.143
3.6 ∞
2
8
2 x x3 (a) E(X) = xf ( x)dx = x dx + x 2 dx = 8 24 x −∞ 0 2
∫
∫
2
∫
8
+ 2 ln x 2 = 1/3 + 2(ln 8 – ln 2) ≅ 3.106 0
3
(b) P(X < 3 | X > 2) =
1− 2/3 = 4/9 1 − 1/ 4
P( X < 3 and X > 2) P(2 < X < 3) = = P ( X > 2) P ( X > 2)
∫ 2
2 dx x2 2
x 1− dx 8 0
∫
−
=
2 x
3
2 2
⎡x ⎤ 1− ⎢ ⎥ ⎣ 16 ⎦ 0 2
=
3.7 (a) The mean and median of X are 13.3 lb/ft2 and 11.9 lb/ft2, respectively (as done in Problem 3-3-3). (b) The event “roof failure in a given year” means that the annual maximum snow load exceeds the design value, i.e. X > 30, whose probability is P(X > 30) = 1 – P(X ≤ 30) = 1 – FX(30) 4 = 1 – [1 – (10/30) ] 4 = (1/3) = 1/81 ≅ 0.0123 ≡ p Now for the first failure to occur in the 5th year, there must be four years of non-failure followed by one failure, and the probability of such an event is 4 4 (1 – p)4p = [1 - (3/4)4 ] ×(1/3) ≅ 0.0117 (b) Among the next 10 years, let Y count the number of years in which failure occurs. Y follows a binomial distribution with n = 10 and p = 1/81, hence the desired probability is P(Y < 2) = P(Y = 0) + P(Y = 1) n n–1 = (1 – p) + n(1 – p) p 10 = (80/81) + 10×(80/81)9×(1/81) ≅ 0.994
3.8 (a) Note that we are given the CDF, FX(x) = P(X ≤ x) and it is a continuous function. Hence P(2 ≤ X ≤ 8) = FX(8) – FX(2) = 0.8 – 0.4 = 0.4 (b) The median is defined by the particular x value where F(x) = 0.5, which occurs somewhere along x = 2 and x = 6; precisely, the constant slope there gives (0.5 – 0.4) / (xm – 2) = (0.6 – 0.4) / (6 – 2) ⇒ xm = 2 + 2 = 4 (c) The PDF is the derivative (i.e. slope) of the CDF, hence it is the following piecewise constant function:
f (x) X
0.2 0.1 0.05
0.05 2
6
10
x
(d) From the CDF, it is seen that one may view X as a discrete R.V. which takes on the values 1, 4, 8 with respective probabilities 0.4, 0.2, 0.4 (recall probability = area of each strip) for calculating E(X). Hence E(X) = 1×0.4 + 4×0.2 + 8×0.4 = 4.4 (integration would give the same result)
3.9 (a) Differentiating the CDF gives the PDF, 0 for s ≤ 0 ⎧ ⎪ s2 s + for 0 < s ≤ 12 f S ( s ) = ⎨288 24 ⎪ 0 for s > 12 ⎩ ~ The mode s is where fS has a maximum, hence setting its derivative to s )=0 fS’( ~ ~ ⇒ - s /144 + 1/24 = 0 ⇒ the mode ~ s = 6. The mean of S, ∞
E(S) =
∫
12
∫
sf ( s ) ds = (
s = −∞
0
=
− s3 s2 + )ds 288 24
− (12 4 ) 12 3 + = = - 18 + 24 = 6 4 × 288 3 × 24
(b) Dividing the sample space into two regions R = 10 and R = 13, the total probability of failure is P(S > R) = P(S > R | R = 10)P(R = 10) + P(S > R)P(R = 13) = [1 – FS(10)]×0.7 + [1 – FS(13)]×0.3 3 2 = [1 – (-10 / 864 + 10 / 48)]×0.7 + [1 – 1]×0.3 = 0.074074074×0.7 ≅ 0.0519
3.10 (a) The only region where fX(x) is non-zero is between x = 0 and x = 20, where fX(x) = F’X(x) = - 0.005x + 0.1 which is a straight line segment decreasing from y = 0.1 (at x = 0) to y = 0 (at x = 20)
fX(x ) x 0
xm
2
(b) The median divides the triangular area under fX into two equal parts, hence, comparing the two similar triangles which have area ratio 2:1, one must have 0.5 (20 – xm) / 20 = (1 / 2) 0.5 ⇒ xm = 20(1 – 0.5 ) ≅ 5.858
3.11
x<0 0< x≤2 ; sketches follow: 2< x≤5 x>5
0 ⎧ ⎪ x 2 / 16 ⎪ (a) Integration of the PDF gives FX ( x) = ⎨ ⎪ x / 4 − 1/4 ⎪⎩ 1
fX(x ) 1/
1
FX(x ) (2,1/
x 0
5
2
x 0 2
2
5
5
⎡ x2 ⎤ ⎡ x3 ⎤ x( x / 8)dx + x(1 / 4)dx = ⎢ ⎥ + ⎢ ⎥ = 71 / 24 ≅ 2.96 (mm) ⎣ 24 ⎦ 0 ⎣ 8 ⎦ 2 0 2 (c) P(X < 4) = 1 – P(X > 4) Where P(X > 4) is easily read off from the PDF as the area (5 – 4)(1/4) = 1/4, hence P(X < 4) = 1 – 1/4 = 3/4 = 0.75 2
(b) E(X) =
∫
5
∫
(d) A vertical line drawn at the median xm would divide the unit area under fX into two equal halves; the right hand rectangle having area 0.5 = (5 – xm)(1/4) ⇒ xm = 5 – 4(0.5) = 3 (mm) (e) Since each of the four cracks has p = 0.25 probability of exceeding 4mm (as calculated in (c)), only one of them exceeding 4mm has the binomial probability (where n = 4, p = 0.25) 3 4×0.75 ×0.25 ≅ 0.422
3.12 x
(a) For 3 ≤ x ≤ 6, FX(x) =
24
∫t
3
dt = 4 / 3 − 12 / x 2 ; elsewhere FX is either 0 (for x < 3) or 1 (for x >
3
6) FX is a parabola going from x = 3 to x = 6, followed by a horizontal line when plotted 6
(b) E(X) =
24
∫ x( x
3
)dx = 24(1/3 – 1/6) = 4 (tons)
3
6
2
2
(c) First calculated the variance, Var(X) = E(X ) – [E(X)] =
∫x
2
(
3
24 )dx - 16 = 24(ln 6 – ln 3) = x3
24 ln 2 - 16 ⇒ c.o.v. =
σ 24 ln 2 − 16 = ≅ 19.9% µ 4
(d) When the load X exceeds 5.5 tons, the roof will collapse, hence P(roof collapse) = P(X > 5.5) = 1 – P(X ≤ 5.5) 2 = 1 – FX(5.5) = 1 – (4/3 – 12/5.5 ) ≅ 0.063
3.13 (a)
Since the return period, τ, is 200 years, this means the yearly probability of exceedance (i.e. encountering waves exceeding the design value) is p = 1 / τ = 1 / (200 years) = 0.005 (probability per year)
(b)
For each year, the probability of non-exceedance is 1 – 0.005 = 0.995, while the intended lifetime is 30 years. Hence non-exceedance during the whole lifetime has the binomial probability, 0.995 ≅ 0.860 30
3.14 Design life = 50 years Mean rate of high intensity earthquake = 1/100 = 0.01/yr P(no damage within 50 years) = 0.99 Let P = probability of damage under a single earthquake (a)
Using a Bernoulli Sequence Model for occurrence of high intensity earthquakes, P(quake each year) = 1/100 = 0.01 P(damage each year) = 0.01p P(no damage in 50 years) = (1 – 0.01p)50 ≡ 0.99 Hence, 1 – 0.01p = (0.99)1/50 = (0.99)0.02 = 0.9998 or p = 0.02
(b)
Assuming a Poisson process for the quake occurrence, Mean rate of damaging earthquake = 0.01p = 0.01x0.02 = 0.0002/yr P(damage in 20 years) = 1 – P(no damage in 20 years) = 1 – P(no occurrence of damaging quakes in 20 years) = 1 – e-0.0002x20 = 1 – e-0.004 = 0.004
3.15 Failure rate = ν = 1/5000 = 0.0002 per hour (a)
P(no failure between inspection) = P(N=0 in 2500 hr) =
(0.0002 × 2500) 0 e −0.0002×2500 = e-0.5 = 0.607 0!
Hence P(failure between inspection) = 1 – 0.607 = 0.393 (b)
P(at most two failed aircrafts among 10) = P(M=0)-P(M=1)-P(M=2) =
10! 10! 10! (0.393)0(0.607)10 (0.393)(0.607)9 (0.393)2(0.607)8 0!× 10! 1!× 9! 2!× 8!
= 0.0068 – 0.044 – 0.1281 = 0.179 (c)
Let t be the revised inspection/maintenance interval P(failure between inspection) = 1-P(no failure between inspection) = 1-e-0.0002t = 0.05 -0.0002t = ln 0.95 and t = 256.5 → 257 hours
3.16 (a) Let C be the number of microscopic cracks along a 20-feet beam. C has a Poisson distribution with mean rate νC = 1/40 (number per foot), and length of observation t = 20 feet, hence the parameter λ = (1/40)(20) = 0.5, thus P(C = 2) = e-0.5 (0.52 / 2!) ≅ 0.076 (b) Let S denote the number of slag inclusions along a 20-feet beam. S has a Poisson distribution with mean rate νS = 1/25 (number per foot), and length of observation t = 20 feet, hence the parameter λ = (1/25)(20) = 0.8, thus 1 – P(S = 0) = 1 - e-0.8 ≅ 0.551 (c) Let X be the total number of flaws along a 20-feet beam. Along 1000 feet (say) of such a beam, one can expect (1/40)(1000) = 25 cracks and (1/25)(1000) = 40 slag inclusions. Hence the mean rate of flaw would be ν = (25 + 40)/1000 or simply (1/40) + (1/25) = 0.065 flaws per foot, which is multiplied to the length of observation, t = 20 feet to get the parameter λ = 1.3. Hence P(X > 2) = 1 – P(X ≤ 2) = 1 – e ≅ 0.143
-1.3
2
(1 + 1.3 + 1.3 / 2!) = 1 - 0.857112489
(d) Thinking of beam rejection as “success”, the total number (N) of beams rejected among 4 would follow a binomial distribution with n = 4 and p = answer in part (c). Thus ⎛ 4⎞ P(N = 1) = ⎜⎜ ⎟⎟ 4×0.142887511×0.8573 ≅ 0.360 ⎝1⎠
3.17 (a) Let N be the number of poor air quality periods during the next 4.5 months; N follows a Poisson process with mean value (1/month)(4.5 months) = 4.5, hence -4.5 2 P(N ≤ 2) = e (1 + 4.5 + 4.5 /2!) ≅ 0.174 (b) Since only 10% of poor quality periods have hazardous levels, the “hazardous” periods (H) must occur at a mean rate of 1 per month×10% = 0.1 per month, hence, over 3 months, H has the mean λH = (0.1)(3) = 0.3 -0.3 ≅ 0.259 ⇒ P(ever hazardous) = 1 – P(H = 0) = 1 - e Alternative approach: use total probability theorem: although there is (1 – 0.1) = 0.9 probability of non-hazardous pollution level during a poor air quality period, during a 3month period there could be any number (n) such periods and the probability of nonn hazardous level reduces to 0.9 for a given n. Hence the total probability of non-hazardous level during the whole time is ∞ ∞ ∞ e − (1×3) (1 × 3) n (3 × 0.9) n -3 3×0.9 -0.3 =e e 0.9 n P ( N = n) = = e , hence 0.9 n = e −3 n ! n ! n =0 n =0 n =0 -0.3 P(ever hazardous) = 1 – P(never hazardous) = 1 – e = 1 - 0.741 ≅ 0.259
∑
∑
∑
3.18 (a)
Let E and T denote the number of earthquakes and tornadoes in one year, respectively. They are both Poisson random variables with respective means 1 λE = νEt = ×1 year = 0.1; λT = 0.3 10 years Also, the (yearly) probability of flooding, P(F) = 1/5 = 0.2, hence, due to statistical independence among E, T, F -0.4 ⇒ P(good) = P(E = 0)P(T = 0)P(F’) = e-0.1 e-0.3 (1 – 0.2) = e ×0.8 ≅ 0.536
Note: alternatively, we can let D be the combined number of earthquakes or/and tornadoes, with mean rate νD = νE +νT = 0.1 + 0.3 = 0.4 (disasters per year), and compute -0.4 P(D = 0)P(F’) = e ×0.8 instead (b)
In each year, P(good year) ≡ p ≅ 0.536 (from (a)). Hence P(2 out of 5 years are good) ⎛ 5⎞ = ⎜⎜ ⎟⎟ p 2 (1 − p) 3 ≅ 0.287 ⎝ 2⎠
(c)
Let’s work with D as defined in (a). P(only one incidence of natural hazard) = P(D = 0)P(F) + P(D = 1)P(F’) –0.4 –0.4 =e ×0.2 + (e ×0.4)×(1 – 0.2) ≅ 0.349
3.19 Mean rate of accident = 1/3 per year (a)
P(N=0 in 5 years) = e-1/3x5 = 0.189
(b)
Mean rate of fatal accident = 1/3x0.05 = 0.0167 per year P(failure between inspection) = 1-P(no fatal accident within 3 years) = 1-e-0.0167x3 = 0.0489
3.20 (a)
Let X be the number of accidents along the 20 miles on a given blizzard day. X has a 1 ×20 miles = 0.4, hence Poisson distribution with λX = 50 miles –0.4 P(X ≥ 1) = 1 – P(X = 0) = 1 – e = 1 – 0 670320046 ≅ 0.33
(b)
Let Y be the number of accident-free days among five blizzard days. With n = 5, and p = daily accident-free probability = P(X = 0) ≅ 0 670, we obtain ⎛ 5⎞ P(Y = 2) = ⎜⎜ ⎟⎟ p 2 (1 − p) 3 ≅ 0.16 ⎝ 2⎠
3.21 (a)
Let X be the number of accidents in two months. X has a Poisson distribution with 3 ×2 months = 0.5, hence λX = 12 months –0.5 P(X = 1) = e × 0.5 ≅ 0.303, whereas –(3/12)(4)
2
[(3/12)(4)] / 2! P(2 accidents in 4 months) = e –1 = e /2! ≅ 0.184 No, P(1 accidents in 2 months) and P(2 accidents in 4 months) are not the same.
(b)
20% of all accidents are fatal, so the mean rate of fatal accidents is νF = νx×0.2 = 0.05 per month Hence the number of fatalities in two months, F has a Poisson distribution with mean λF = (0.05 per month)(2 months) = 0.1, hence –0.1 P(fatalities in two months) = 1 – P(F = 0) = 1 – e ≅ 0.095
3.22 The return period τ (in years) is defined by τ =
1 where p is the probability of flooding per year. p
Therefore, the design periods of A and B being τA = 5 and τB = 10 years mean that the respective yearly flood probabilities are P(A) = 1/ (5 years) = 0.2 (probability per year) P(B) = 1/(10 years) = 0.1 (probability per year) (a) P(town encounters any flooding in a given year) = P(A∪B) = P(A) + P(B) - P(AB) = P(A) + P(B) - P(A)P(B) (∵ A and B s.i.) =
1
τA
+
1
τB
−
1
(i)
τ Aτ B
= 0.2 + 0.1 - 0.2⋅0.1 = 0.28 (b) Let X be the total number of flooded years among the next half decade. X is a binomial random variable with parameters n = 5 and p = 0.28. Hence P(X ≥ 2) = 1 - P(X < 2) = 1 - f(0) - f(1) = 1 - (1 - 0.28)5 - 5×(0.28)×(1 - 0.28)4 ≈ 0.43 (c) Let τA and τB be the improved return periods for levees A and B, respectively. Using (i) from part (a), we construct the following table: New τA (and cost)
New τB (and cost)
10 ($5M) 10 ($5M) 20 ($20M) 20 ($20M)
20 ($10M) 30 ($20M) 20 ($10M) 30 ($20M)
Yearly flooding probability 1 1 1 = + − τ A τ B τ Aτ B 0.145 0.130 0.0975 0.0817
Total cost (in million dollars) 5 + 10 = 15 5 + 20 = 25 20 + 10 = 30 20 + 20 = 40
Since the goal is to reduce the yearly flooding probability to at most 0.15, all these options will work but the top one is least expensive. Hence the optimal course of action is to change the return periods of A and B to 10 and 20 years, respectively.
3.23 Given: ν = mean flaw rate = 1 (flaw) / 50m2; t = area of a panel = 3m x 5m = 15m2 Let X be the number of flaws found on area t. X is Poisson distributed, i.e. f(x) = e
−λ
λx x!
with λ = νt = 15 / 50 = 0.3 (flaws) (a)
P(replacement) = 1 - P(0 or 1 flaw) = 1 - f(0) - f(1) = 1 - e-0.3 - 0.3 e-0.3 ≈ 0.037
(b) Since the probability of replacement is 0.037 and there are 100 panels, we would expect 0.037×100 = 3.7 replacements on average, which gives the expected replacement cost 3.7×$5000 ≅ $18500 (c) For the higher-grade glass, ν = 1 (flaw) / 80m2 ⇒ λ = 15/80 = 0.1875, with which we can calculate the new probability P(replacement) = 1 - e-0.1875 - 0.1875 e-0.1875 ≈ 0.0155, which gives rise to an expected replacement cost of 100×0.0155×$5100 ≈ $7920 (assuming each higher grade panel is also $100 more expensive to replace) We can now compare the total costs: Let C = initial cost of old type panels Old type: Expected total cost = C + Expected replacement cost = C + $18500 New type: Expected total cost = C + Extra initial cost + Expected replacement cost = C + $100×100 + $7920 = C + $17920, which is less than that of the old type, ∴ the higher grade panels are recommended.
3.24 (a) Let X be the number of trucks arriving in a 5-minute period. Given: truck arrival has mean rate ν = 1 (truck) / minute; hence with t = 5 minutes ⇒ λ = νt = 5 (trucks) Hence P(X ≥ 2) = 1 - P(X < 2) = 1 - P(X = 0) - P(X = 1) = 1 - e-5 - 5e-5 ≈ 0.96 (b) Given: P(a truck overloads) ≡ p = 0.1 ; number of trucks ≡ n = 5. We can use the binomial distribution to compute P(at most 1 truck overloaded among 5) = b(0; n, p) + b(1; n, p) = (1 - 0.1)5 + 5⋅(1 - 0.1)4⋅(1 - 0.1) ≈ 0.92 (c) We will first find probability of the compliment event. During t = 30 minutes, the average number of trucks is λ = νt = 1×30 = 30 (trucks). Any number (x) of trucks could pass, with the (Poisson) probability f(x) = λx e-λ / x! but if none of them is to be overloaded (event NO), the (binomial) probability is P(NO | x trucks pass) = 0.9x Hence the total probability of having no overloaded trucks during 30 minutes is P(NO | x = 0)P(x = 0) + P(NO | x = 1)P(x = 1) + …… x ∞ x − 30 30 = ∑ 0.9 e x! x =0 ∞
= e-30
27 x ∑ x = 0 x!
= e-30 e27 = e-3 , hence the desired probability –3 P(any overloaded trucks passing) = 1 – e ≅ 0.95
3.25 (a) Let X be the number of tornadoes next year. X has a Poisson distribution with 20 occurences = 1 (per year); ν= 20 years t = 1 year ⇒ λ = νt = 1, hence
P(next year will be a tornado year) = P(X ≥ 1) = 1 – P(X = 0) -λ = 1 – e = 1 – 1/e ≅ 0.632 (b) This is a binomial problem, with n = 3 and p = yearly probability of having any tornado(es) = 0.6321. Hence P(two tornado years in three years)
⎛ 3⎞ ⎝2⎠
= ⎜ ⎟ 0.632 2 (1 − 0.632)1 ≅ 0.441
(c) (i)
(ii)
The tornadoes follow a Poisson model with mean rate of occurrence ν = 1 (per year), hence for an observation period of t = 10 (years), the expected number of tornadoes is λ = νt = 10 Number of tornado years follows a binomial model with n = 10, p = 0.632120559, hence the mean (i.e. expected) number of tornado years over a ten-year period is np ≅ 6.32
3.26 (a) Let X be the number of strong earthquakes occurring within the next 20 years. X has a Poisson distribution with λ = (1/50 years)(20 years) = 0.4, hence P(X ≤ 1) = P(X = 0) + P(X = 1) = e-0.4 (1 + 0.4) ≅ 0.938 (b) Let Y be the total number (maximum 3) of bridges that will collapse during a strong earthquake; Y has a binomial distribution with parameters n = 3, p = 0.3, hence the required probability is P(Y = 1) = 3×0.3×(1 – 0.3)2 = 0.441 (c) During a strong earthquake, the probability that all 3 bridges survive is (1 – 0.3)3 = 0.73 = 0.343. But since any number of strong earthquakes could happen during the next 20 years, we need to compute the total probability that there is no bridge collapse (with X as defined in part (a)), ∞
P(no bridge collapses) =
∑
P(no bridge collapses | x earthquakes occur in 20 years)P(X = x)
x =0
∞
=
∑
(0.343)x [(e-0.4)(0.4x) / x!]
x =0
∞
= (e-0.4)
∑
[(0.343)(0.4)]x / x!
x =0
= (e-0.4)(e0.1372) = e-0.2628 ≅ 0.769
3.27 (a)
Let X be the total number of excavations along the pipeline over the next year; X has a Poisson distribution with mean λ = (1/50 miles)(100 miles) = 2, hence P(at least two excavations) = 1 – P(X = 0) – P(X = 1) = 1 – e-λ (1 + λ) = 1 – e-2(3) ≅ 0.594
(b) For each excavation that takes place, the pipeline has 0.4 probability of getting damaged, and hence (1 – 0.4) = 0.6 probability of having no damage. Hence P(any damage to pipeline | X = 2) = 1 – P(no damage | X = 2) = 1 – 0.62 = 1 – 0.36 = 0.64 Alternative method: Let Di denote “damage to pipeline in i-th excavation”; the desired probability is P(D1∪ D2) = P(D1) + P(D2) – P(D1D2) = P(D1) + P(D2) – P(D1 | D2) P(D2) = P(D1) + P(D2) – P(D1 ) P(D2) = 0.4 + 0.4 – 0.42 = 0.8 – 0.16
= 0.64 (c) Any number (x) of excavations could take place, but there must be no damage no matter what x value, hence we have the total probability ∞
∑
P(no damage | x excavations) P(x excavations)
x =0
∞
=
∑
e −λ λ x x! (0.6λ ) x -λ 0.6λ =e e x!
x
0.6
x =0
∞
= e-λ ∑ x =0
-0.4λ
-0.4(2)
=e =e ≅ 0.449
-0.8
=e
Alternative method: recall the meaning of ν in a Poisson process—it is the mean rate, i.e. the true proportion of occurrence over a large period of observation. Experimentally, it would be determined by
ν=
nE N
where nE is the number of excavations observed over a very large number (N) of miles of pipeline. Since 40% of all excavations are damaging ones, damaging excavations must also
Hence P(no damaging excavation over a 100 mile pipeline) –(1/125 mi.)(100 mi.) = e–100/125 = e–0.8 =e ≅ 0.449
3.28 (a)
Let X be the number of flaws along the 30-inch weld connection. X has a Poisson distribution with 0 .1 λX = ×30 inches = 0.25, hence 12 inches –0.25 P(acceptable connection) = P(X = 0) = e ≅ 0.779
(b)
Let Y be the number of acceptable connections among three. Y has a binomial distribution with n = 3 and p = P(an acceptable connection) = 0.778800783, hence the desired probability is P(Y ≥ 2) = P(Y = 2) + P(Y = 3) = 3 p 2 (1 − p ) + p 3 ≅ 0.875
(c)
The number of flaws (F) among 90 inches of weld has a Poisson distribution with 0 .1 λF = ×90 inches = 0.75, hence 12 inches Hence P(1 flaw in 3 connections) = P(1 flaw in 90 inches of weld) = P(F = 1) –0.75 =e ×0.75 ≅ 0.354 Note: “1 flaw in 3 connections” is not the same as “1 unacceptable connections among 3”, as an unacceptable connection does not necessarily contain only 1 flaw—it could have 2, 3, 4, etc.
3.29 Mean rate of flood occurrence = 6/10 = 0.6 per year (a)
P(1 ≤ N ≤ 3 within 3 years) =[
(0.6 × 3) 1 (0.6 × 3) 2 (0.6 × 3) 3 − 0.6×3 + + ]e 1! 2! 3!
= (1.8+1.62+0.972)x0.165 = 0.725 (b)
Mean rate of flood that would cause inundation = 0.6x0.02 P(treatment plant will not be inundated within 5 years) = P(no occurrence of inundating flood) = e-0.02 = 0.98
3.30 Given: both I and N are Poisson processes, with νI = 0.01/mi and νN = 0.05/mi. Hence, along a 50-mile section of the highway, I and N have Poisson distributions with respective means λI = (0.01/mi)(50 mi) = 0.5, λN= (0.05/mi)(50 mi) = 2.5 (a)
P(N = 2) = e − λ N (λN )2 / 2! = e–2.5 2.52 / 2 ≅ 0.257
(b)
A, accident of either type, has the combined mean rate of occurrence νA = νI + νN = 0.06, hence A has a Poison distribution with λA = (0.06/mi)(50 mi) = 3, thus P(A ≥ 3) = 1 – P(A = 0) – P(A = 1) – P(A = 2) –3 2 = 1 – e (1 + 3 + 3 /2) ≅ 0.577
(c)
Let us model this as an n = 2 binomial problem, where each “trial” corresponds to an accident, in which “success” means “injury” and “failure” means “non-injury”. How do we determine p? Consider the physical meaning of the ratio νI : νN = 0.01 : 0.05 -- it compares how frequently I occurs relative to N, hence their relative likelihood must be 1 to 5, thus P(N) = 5/6 and P(I) = 1/6 whenever an accident occurs. Therefore, p = 5/6, and 2 2 P(2 successes among 2 trials) = p = (1/6) ≅ 0.028
3.31 (a)
Let W and S denote the number of winter and summer thunderstorms, respectively. Their mean occurrence rates (in numbers per month) are estimated based on the observed data as (i) νW ≅ (173)/(21×6) ≅ 1.37; (ii) νS ≅ (840)/(21×6) = 6.67
(b)
Let us “superimpose” the two months as one, with winter and summer thunderstorms (X) that could happen simultaneously, at a combined mean rate of νX = (173 + 840)/(21×6) = 8.03968254 (storms per month), thus λX = (νX )(1 month) = 8.03968254, hence -8.040 4 P(X = 4|t = 1 month) = e (8.040 )/4! ≅ 0.056
(c)
In any particular year, the (Poisson) probability of having no thunderstorm in December is -173 / 126 = 0.253, P(W = 0 | t = 1 month) = e let’s call this p. Having such a “success” in 2 out of the next 5 years has the (binomial) probability ⎛ 5⎞ 2 ⎜⎜ ⎟⎟ p (1 – p)3 = 5!/2!/3!× 0.2532×(1 - 0.253)3 ≅ 0.267 ⎝ 2⎠
3.32 (a)
(b)
(c)
Let D denote defects and R denote defects that remain after inspection. Since only 20% of defects remain after inspection, we have 1 νR = νD×0.2 = ×0.2 = 0.001 per meter 200 meters For t = 3000 meters of seams, undetected defects have a mean of λ = νR t = (0.001/m)(3000m) = 3, ⇒ P(more than two defects) = 1 – P(two or less defects) = 1 – [P(D = 0) – P(D = 1) – P(D = 2)] –3 2 = 1 – [e (1 + 3 + 3 /2!)] = 1 – 0.423 ≅ 0.577
Suppose the allowable fraction of undetected defects is p (0 < p < 1), then the mean rate of undetected defects is νU = νD×p, hence, for 1000 meters of seams, the (Poisson) mean number of defects is νU×1000m = (1/200m)(p)(1000m) = 5p, thus if we require P(0 defects) = 0.95 –5p ⇒ e = 0.95 ⇒ –5p = ln(0.95) ⇒ p = – ln(0.95)/5 = 0.0103, i.e. only about 1% of defects can go on undetected.
3.33 Let J1 and J2 denote the events that John’s scheduled connection time is 1 and 2 hours, respectively, where P(J1) = 0.3 and P(J2) = 0.7. Also, let X be the delay time of the flight in hours. Note that P(X > x) = 1 – P(X ≤ x) = 1 – F(x) = 1 – (1 – e -x/0.5 =e
-x/0.5
)
(a) Let M denote the event that John misses his connection, i.e. the flight delay time exceeded his scheduled time for connection. Using the theorem of total probability, P(M) = P(M | J1)P(J1) + P(M | J2)P(J2) = P(X > 1)×0.3 + P(X > 2)×0.7 = e– 1/0.5 × 0.3 + e– 2/0.5 × 0.7 = 0.135×0.3 + 0.0183×0.7 ≅ 0.053 (b) Regardless of whether John has a connection time of 1 hour and Mike has 2, or the opposite, for them to both miss their connections the flight must experience a delay of more than two hours, and the probability of such an event is P(X > 2) = e
– 2/0.5
≅ 0.018
(c) Since Mary has already waited for 30 minutes, the flight will have a delay time of at least 0.5 hours when it arrives. Hence the desired probability is P(X > 1 | X > 0.5) = P(X > 1 and X > 0.5) / P(X > 0.5) = P(X > 1) / P(X > 0.5) –1/0.5 –0.5/0.5 –2 –1 =e /e = e / e = 1/e ≅ 0.368
3.34 1000 rebars delivered with 2% below specification (a) 20 rebars are tested; hence, N = 1000, m = 20, n = 20 P(all 20 bars will pass test) = P(zero bars will fail test)
⎛ 20 ⎞⎛ 980 ⎞ ⎜⎜ ⎟⎟⎜⎜ ⎟⎟ ⎝ 0 ⎠⎝ 20 ⎠ = = 0.665 ⎛1000 ⎞ ⎜⎜ ⎟⎟ ⎝ 20 ⎠ (b) P(at least two bars will fail) = 1 – P(at most 1 bar will fail) = 1 – P(x=0) – P(x=1)
⎛ 20 ⎞⎛ 980 ⎞ ⎜⎜ ⎟⎟⎜⎜ ⎟⎟ ⎝ 1 ⎠⎝ 19 ⎠ = 1 – 0.665 – ⎛1000 ⎞ ⎜⎜ ⎟⎟ ⎝ 20 ⎠ = 1 – 0.665 – 0.277 = 0.058 (c) Let n be the required number of bars tested P(all n bars will pass test)
⎛ n ⎞⎛ 980 ⎞ ⎜⎜ ⎟⎟⎜⎜ ⎟⎟ ⎝ 0 ⎠⎝ n ⎠ ≡ 0.10 = ⎛1000 ⎞ ⎜⎜ ⎟⎟ ⎝ n ⎠ n ≅ 107.712 ⇒ 108 bars required
3.35
(a) “A gap larger than 15 seconds” means that there is sufficient time (15 seconds or more) between the arrival of two successive cars. Let T be the (random) time (in seconds) between successive cars, which is exponentially distributed with a mean of E(T) = (1/10) minute = 0.1 minute = 6 seconds, –15/6 –2.5 hence P(T > 15) = 1 – P(T ≤ 15) = e =e ≅ 0.082
(b) For any one gap, there is (1 – e–2.5) chance that it is not long enough for the driver to cross. –2.5 3
For such an event to happen 3 times consecutively, the probability is (1 – e ) ; followed by –2.5 a long enough gap that allows crossing, which has probability e . Hence the desired probability is –2.5 3 –2.5 (1 – e ) (e ) ≅ 0.063 More formally, if N is the number of gaps one must wait until the first “success” (i.e. being able to cross), G follows a geometric distribution, n–1 –2.5 P(N = n) = (1 – p) p where p = e (c) Since the mean of a geometric distribution is 1/p (see Ang & Tang Table 5.1), he has to wait a mean number of –2.5 2.5 1/p = 1/ e =e ≅ 12.18 gaps before being able to cross. (d) First let us find the probability of the compliment event, i.e. P(none of the four gaps were large enough) –2.5 4 = (1 – e )
Hence ) = 1 – 0.70992075 ≅ 0.290
–2.5 4
P(cross within the first 4 gaps) = 1 – (1 – e
3.36 Standard deviation of crack length= 6.25 = 2.5 (a) Let X be the length (in micrometers) of any given crack.. ⎛ 74 − 71 ⎞ = 1 − Φ 1.2 = 1.2 P(X > 74) = 1 − Φ ⎜ ( ) ⎟ ⎝ 2.5 ⎠
(b) Given that 72 < X , the conditional probability P(X > 77⏐ X > 72) = P(X > 77 and X > 72) / P(X > 72) = P(X > 77) / P(X > 72) = [1 - Φ((77 - 71)/2.5)] / [1 - Φ((72 - 71)/2.5)] = [ 1 - Φ(2.4)] / [1 - Φ(0.4)] =0.0082 / 0.8446 ≅ 0.024
3.37 (a)
P(activity C will start on schedule) = P(A<60) ∩ P(B<60) 60 − 50 60 − 45 = Φ( )Φ ( ) = Φ (1)Φ (1) 10 15 = 0.8412 = 0.707
(b)
P(Project completed on target) = P(T) = P(T⏐E)P(E) + P(T⏐ E )P( E ) = P(C<90)x0.707 + P(C<90-15)x0.293 90 − 80 ) × 0.707 + Φ ( −0.2) × 0.293 = Φ( 25 = 0.655x0.707 + 0.421x0.293 = 0.586
3.38 Let, T = The present traffic volume = N(200,60) (a) P(T>350) = 1 - Φ((350 - 200)/60) = 1 - Φ(2.5) = 0.00621 (b) Let, T1 = Traffic volume after 10 years. µ T1 = 200 + 0.10×200×10 = 400
δ T = δ T = 60/200 = 0.3 1
Hence,
σ T = 0.3×400 = 120 1
So, T1 is N(400, 120) P(T1>350) = 1 - Φ((350 - 400)/120) = 1 - Φ(0.42) = 0.662 (c) C = Capacity of airport after 10 years. P(T1>C) = 0.00621 or, 1 - Φ((C - 400)/120) = 1 – 0.99379 = 1- Φ(2.5) or, (C - 400)/120 = 2.5 or, C = 300 + 400 = 700
3.39 A = volume of air traffic C = event of overcrowded (a)
P(C⏐N-S) = P(A>120) 120 − 100 ) = 1 − Φ (2) = 0.02275 = 1 − Φ( 10
(b)
P(C⏐E-W) = P(A>115) 115 − 100 ) = 1 − Φ (1.5) = 0.0668 = 1 − Φ( 10
(c)
P(C) = P(C⏐N-S)P(N-S)+P(C⏐E-W)P(E-W) = 0.02275x0.8 + 0.0668x0.2 = 0.0315
3.40 Let X be the settlement of the proposed structure; X ~ N( µ X , σ X ). Given probability: P(X ≤ 2) = 0.95 Also, since we’re given the c.o.v. = Hence,
P( X ≤ 2) = Φ ( or
and
σX = 0.2 , σ X = 0.2 µ x µX
2 − µx ) = 0.95 0.2 µ x 2 − µx = Φ −1 (0.95) = 0.95 = 1.645 0.2 µ x
µ x = 1.5 P( X > 2.5) = 1 − Φ (
2.5 − 1.5 ) = 1 − Φ (3.063) = 0.00047 0.2 × 1.5
3.41 (a) Let X be her cylinder’s strength in kips. To be the second place winner, X must be above 70 but below 100, hence P(70 < X < 100) 100 − 80 70 − 80 = Φ( ) − Φ( ) 20 20 = Φ (1) − Φ ( −0.5)
= 0.841 − 0.309 = 0.532 (b) P(X > 100 | X > 90) = P(X > 100 and X > 90) / P(X > 90) = P(X > 100) / P(X > 90) = {1 – Φ[(100 – 80)/20]} / {1 – Φ[(90 – 80)/20]} = [1 – Φ(1)] / [1 – Φ(0.5)] = 0.159 / 0.309 ≅ 0.514 (c) Let Y be the boyfriend’s cylinder strength in kips, which has a mean of 1.01×80 = 80.8. Therefore, Y ~ N(80.8, σY) Let D = Y – X; D is normally distributed with a mean of µD = µY – µX = 80.8 – 80 = 0.8 > 0 This suffices to conclude that the guy’s cylinder is more likely to score higher. Mathematically, 0 − 0.8 P(D > 0) = Φ ( ) > 0.5
σD
hence it is more likely for the guy’s cylinder strength to be higher than the girl’s.
3.42 H = annual maximum wave height = N(4,3.2)
6−4
(a)
P(H>6) = 1 − Φ (
(b)
Design requirement is P(no exceedance over 3 years) = 0.8 = (1-p)3 Where p = p(no exceedance in a given year)
3.2
) = 1 − Φ (0.625) = 0.266
Hence, p = 1-(0.8)1/3 = 0.0717 That is, Φ (
h−4
where h is the designed wave height ) = 0.0717 3.2 h = 3.2 Φ -1(0.0717) + 4 = 3.2x0.374 + 4 = 5.2 m
(c)
Probability of damaging wave height in a year = 0.266x0.4 = 0.1064 Hence mean rate of damaging wave height = 0.1064 per year P(no damage in 3 years) = P(no damaging wave in 3 years) = e-0.1064x3 =0.727
3.43 Let X be the daily SO2 concentration in ppm, where X ~ N(0.03, 0.4×0.03), i.e., X is normal with mean 0.03 and standard deviation 0.012. Thus the weekly mean, X ~ N(0.03, 0.012/ 7 ). Hence 0.04 − 0.03 ⎞ 1. P( X < 0.04) = Φ ⎛⎜ ⎟ = Φ ( 2.205) = 0.986 , hence ⎝ 0.012 / 7 ⎠ P1(violation) = 1 – 0.9863 = 0.0137, whereas
2. On any one day, the probability of X being under 0.075 is 0.075 − 0.03 ⎞ P(X < 0.075) = Φ ⎛⎜ ⎟ = Φ( 3.75) ≅, hence 0.012 ⎠ ⎝ P(no violation) = P(0 or 1 day being under 0.075 out of 7 days) 7 6 = 1 + 7×1 ×(1 – 1) =1 ⇒ P2(violation) = 1 – 1 =0 < P1(violation), ∴criteria 1 is more likely to be violated, i.e. more strict.
3.44 Let F be the daily flow rate; we're given F ~ N(10,2) (a) P(excessive flow rate) = P(F > 14) ⎛ 14 − 10 ⎞ = 1− Φ⎜ ⎟ ⎝ 2 ⎠
= 1 − Φ ( 2)
= 1 − 0.97725 ≈ 0.02275 (b) Let X be the total number of days with excessive flow rate during a three-day period. X follows a binomial distribution with n = 3 and p = 0.02275 (probability of excessive flow on any given day), hence P(no violation) = P(zero violations for 3 days) = P(X = 0) = (1 - p)3 ≅ 0.933 (c) Now, with n changed to 5, while p = 0.02275 remains the same, P(not charged) = P(X = 0 or X = 1) = P(X = 0) + P(X = 1) = (1 - p)5 + 5⋅p⋅(1 - p)4 ≅ 0.995 which is larger than the answer in (b). Since the non-violation probability is larger, this is a better option. (d) In this case we work backwards—fix the probability of violation, and determine the required parameter values. We want P(violation) = 0.01 ⇒ P(non-violation) = 0.99 ⇒ (1 - p)3 = 0.99 ⇒ p = 0.00334, but recall from part (a) that p is obtained by computing P(F > 14) ⇒ 0.003344 =1 − Φ ⎜⎛ 14 − µ F ⎟⎞ ⎝ σF ⎠ hence,
14 − µ F =Φ -1 ( 0.99666 ) = 2.712 ⇒ µ F ≅ 8.58 2
3.45 (a) The "parameters" λ and ζ of a log-normal R.V. are related to its mean and standard deviation µ and σ as follows:
⎛σ ⎞ ζ = ln(1 + ⎜ ⎟ ) ⎝ µ⎠ 2
2
λ = ln µ −
ζ2
2 σT Substituting the given values δT ≡ = 0.4, µT = 80 , we find µT
ζ 2 = 0.14842 and λ = 4.307817, hence ζ ≈ 0.385 and λ ≈ 4.308 The importance of these parameters is that they are the standard deviation and mean of the related variable X ≡ ln(T), while X is normal. Probability calculations concerning T can be done through X, as follows:
⎛ ln 20 − 4.3078 ⎞ = Φ ( −3.406) = 0.00033 ⎟ 0.385 ⎝ ⎠
P (T ≤ 20) = Φ ⎜
(b)
(c) "T > 100" is the given event, while "a severe quake occurs over the next year" is the event "100 < T < 101", hence we seek the conditional probability P(100 < T < 101 | T > 100) = P(100 < T < 101 and T > 100) / P(T > 100) = P(100 < T < 101) / P(T > 100) = P(ln 100 < ln T < ln 101) / P(ln T > ln 100) ⎛ ln101 − λ ⎞ ⎛ ln100 − λ ⎞ Φ⎜ −Φ⎜ ⎟ ⎟ ζ ζ ⎠ ⎝ ⎠ = ⎝ ⎛ ln100 − λ ⎞ 1− Φ⎜ ⎟
⎝
=
ζ
0.787468 − 0.779895
1 − 0.779895 = 0.034
⎠
3.46 (a) Let X be the daily average pollutant concentration. The parameters of X are:
ζ ≅ δ = 0.2, λ ≅ ln 60 – 0.22 / 2 ≅ 4.074 Hence P(X > 100) = 1 – P(X ≤ 100) = 1 – Φ( ln100 − 4.074 ) 0.2 = 1 – Φ(2.654) = 1 – 0.996024 ≅ 0.004 (b) Let Y be the total number of days on which critical level is reached during a given week. Y follows a binomial distribution with parameters n = 7, p = 0.004 (and 1 – p = 0.9964), hence the required probability is P(Y = 0) = 0.9967 ≅ 0.972
3.47
ζ = ln(1 + δ 2 ) = 0.2936 H 1 2
λ = ln µ − ζ 2 = 2.953 (a)
P(25 m long pile will not anchor satisfactorily) = P(H>25-2) = P(H>23) = 1 − Φ(
(b)
P(H>2T⏐H>24) =
ln 23 − 2.953 ) = 1 − Φ (0.622) = 0.267 0.2936
P( H > 27 ∩ H > 24) P( H > 27) = P( H > 24) P( H > 24)
ln 27 − 2.953 ) 1 − Φ (1.677) 0.0468 0.2936 = = = 0.21 = ln 24 − 2.953 1 − Φ( ) 1 − Φ (0.767) 0.222 0.2936 1 − Φ(
3.48 (a) Let X be the pile capacity in tons. X is log-normal with parameters ζX ≅ δX = 0.2, λX ≅ ln 100 – 0.22 / 2 ⎛ ln100 − 4.585 ⎞ = 1 − Φ 0.1 = 1 − 0.54 = 0.46 P(X > 100) = 1 - P(X ≤ 100) = 1 − Φ ⎜ ( ) ⎟ 0.2 ⎝ ⎠
(b) Let L be the maximum load applied; L is log-normal with parameters δL = 15/50 = 0.3 2 1/2 ⇒ ζL = [ln(1 + 0.3 )] = 0.294 ⇒ λL = ln 50 – 0.2942 / 2 = 3.869 It is convenient to formulate P(failure) as P(X < L) = P(X / L < 1) = P(ln(X/L) < ln 1) = P(ln X – ln L < 0) but D = ln X – ln L is the difference of two normals, so it is again normal, with 2 2 1/2 µD = λX – λL = 0.716 and σD = (ζX + ζL ) = 0.355 ⎛ 0 − 0.716 ⎞ = Φ −2.017 = 0.0219 ⇒ P(D < 0) = Φ ⎜ ( ) ⎟ ⎝ 0.355 ⎠ (c) P(X > 100 | X > 75) = P(X > 100 and X > 75) / P(X > 75) = P(X > 100) / P(X > 75) ln 75 − 4.585 )] = [answer to (a)] / [1 – Φ( 0.2 = 0.46 / [1 – Φ(–1.338)] = 0.46 / Φ(1.338) = 0.46 / 0.9096 ≅ 0.506
(d) P(X > 100 | X > 90) = P(X > 100 and X > 90) / P(X > 90) = P(X > 100) / P(X > 90) ln 90 − 4.585 = [answer to (a)] / [1 – Φ( )] 0.2 = 0.460172104 / [1 – Φ(– 0.427)] = 0.46 / Φ(0.427) = 0.46 / 0.665 ≅ 0.692
3.49
⎛σ ⎞ ζ = ln(1 + ⎜ ⎟ ) ⎝ µ⎠ 2
2
λ = ln µ −
ζ2
2 (a) Let X be the maximum wind velocity (in mph) at the given city. X ~ LN(λ, ζ) where
ζ = ln(1 + δ 2 ) ≅ δ = 0.2, thus
λ = ln µ −
ζ2
≅ ln (90) – 0.22 / 2
2 Since ζ and λ are the standard deviation and mean of the normal variate ln(X), ⎛ ln120 − 4.48 ⎞ = 1 – Φ(1.538) = 1 – 0.938≅ 0.062 P(X > 120) = 1 − Φ ⎜ ⎟ 0.2 ⎝ ⎠ (b) To design for 100-year wind means the yearly probability of exceeding the design speed (V) is 1/100 = 0.01, i.e. P(X > V) = 0.01, i.e. P(X ≤ V) = 0.99 ⎛ ln V − 4.48 ⎞ = 0.99 Φ⎜ ⎟ 0.2 ⎝ ⎠ ln V − 4.48
= Φ −1 ( 0.99 ) = 2.33
0.2 V = 141( mph )
3.50 T = time until breakdown = gamma with mean of 40 days and standard deviation of 10 days (a)
k/ν = 40,
c.o.v. = 0.25 =
Hence, k = 16, ν = 0.4
1 k
Let t be the required maintenance schedule interval Hence, P(T
or
Iu(0.4 t, 16) ≡ 0.95
By trial and error, t 40 50 60 58 57
Probability 0.533 0.8435 0.9656 0.9520 0.944
Hence, t ≅ 58 days (b)
P(T < 58+7⏐T > 58) = 1−
(c)
I (0.4 × 65,16) I (26,16) P(T > 65) 0.01417 = 1− u = 1− u = 1− = 0 .7 P(T > 58) I u (0.4 × 58,16) I u (23.2,16) 0.048
P(at least one out of three will breakdown) = 1 – P(none of the three will breakdown) = 1 – (0.95)3 = 0.143
3.51 The capacity C is gamma distributed with a mean of 2,500 tone and a c.o.v. of 35% k/ν = 2500,
(a)
(b)
1
= 0.35
k
Hence, k = 8.16, ν = 0.00327 P(C >3000⏐C > 1500) P (C > 3000) 1 − I u (0.00327 × 3000, 8.16) 1 − I u (9.81, 8.16) 0.745 = = = = 0.838 = P (C > 1500) 1 − I u (0.00327 × 1500, 8.16) I u (4.905, 8.16) 0.889
P(C<2000) = Iu (0.00327x2000, 8.16) = 0.312 Mean rate of damaging quake = 0.312 x 1/20 = 0.0156 P(no damage over 50 years) = e-0.0156x50 = 0.458
(c)
P(at least 4 buildings damaged) = 1 – P(none of the the 4 will be damaged) = 1 – (0.688)4 = 0.776
3.52 Ti = time for loading/unloading operation of ith ship T = waiting time of the merchant ship (a)
Given that there are already 2 ships in the queue, P(T>5) = P(T>5⏐T1 = 2)P(T1 = 2) + P(T>5⏐T1 = 3)P(T1 = 3) = P(T2>3)P(T1 = 2) + P(T2>2)P(T1 = 2) = 0x1/4 + 1/4x3/4 = 0.1875
(b)
P(T>5) = P(T>5⏐N=0)P(N=0) + P(T>5⏐N=1)P(N=1) + P(T>5⏐N=2)P(N=2) + P(T>5⏐N=3)P(N=3) Given 0 ship in the queue, P(T>5) = 0 Given 1 ship in the queue, P(T>5) = P(T1>5) = 0 Given 2 ships in the queue, P(T>5) = 0.1875 from above Given 3 ships in the queue, the waiting time will be at least 6 days. Hence it is certain that P(T>5) = 1 In summary, P(T>5) = 0x0.1 + 0x0.3 + 0.1875x0.4 + 1x0.2 = 0.275
3.53 Traffic volume= V = Beta between 600 and 1100 vph with mean of 750 vph and c.o.v. of 0.20 A = accidenct (a)
P(Jamming occurs on the bridge) = P(J) = P(J⏐A)P(A)+P(J⏐ A )P( A ) = 1x0.02 + P(V>1000)x0.98 For the parameters of Beta distribution for V,
q (1100-600) q+r qr (0.2x750)2 = (1100-600)2 2 (q + r ) (q + r + 1) Hence, r ≈ 0.95,q ≈ 0.41 750 = 600+
P(V>1000) = 1- β u (0.41, 0.95) But u = (1000-600)/(1100-600) = 0.8 P(V>1000) = 1- β 0.8 (0.41, 0.95) = 0.097 P(J) = 0.02 + 0.097x0.98 = 0.971
3.54 By using the binomial distribution, P(2 out of 8 students will fail)
⎛8⎞
= ⎜⎜ ⎟⎟(0.2 ) (0.8) 2 2
6
⎝ ⎠
= 28x(0.2)2(0.8)6 = 0.294 By using the hyper geometric distribution, we need number of passing students in a class of 30 = 24 number of failing students in a class of 30 = 6 Hence P(2 out of 8 students will fail)
⎛ 24 ⎞⎛ 6 ⎞ ⎜⎜ ⎟⎟⎜⎜ ⎟⎟ 28 × 15 ⎝ 6 ⎠⎝ 2 ⎠ = = 5852925 ⎛ 30 ⎞ ⎜⎜ ⎟⎟ ⎝8⎠ =0.00007
3.55 (a)
T = trouble-free operational time follows a gamma distribution with mean of 35 days and c.o.v. of 0.25 k/ν = 35,
1
= 0.25
k
Hence, k = 16, ν = 0.457 P(T>40) = 1-Iu(0.457x40, 16) = Iu(18.3, 16) = 0.736 (b)
For the total of 50 road graders, number of graders, Number of graders with T<40 = 0.1x50 = 5 P(2 among 10 graders selected will have T<40)
⎛ 5 ⎞ ⎛ 45 ⎞ ⎜2⎟⎜ 8 ⎟ 10 × 0.0216 × 109 = 0.021 = ⎝ ⎠⎝ ⎠ = 1.027 × 1010 ⎛ 50 ⎞ ⎜ 10 ⎟ ⎝ ⎠
3.56 Seismic capacity C is lognormal with median of 6.5 and standard deviation of 1.5
λ
= ln 6.5
For lognormal distribution, µ > x m
σ / µ < σ / x m = 1.5 / 6.5 = 0.23 and ξ ≈δ Since λ =ln 6.5 = ln µ − (1.5 / µ ) 2 µ ≈ 6.7; and ξ ≈ 1.5 / 6.7 = 0.198 By trial and error, hence
(a)
(b)
(c)
P(Damage) = P(C<5.5) = Φ (
ln 5.5 − ln 6.5 ) = Φ (−0.84) = 0.2 0.198
P(C > 4 ∩ C > 5.5) P(C > 5.5) = P(C > 4) P (C > 4) ln 5.5 − ln 6.5 1 − Φ( ) 0.907 0.198 = = = 0.915 ln 4 − ln 6.5 0.9929 1 − Φ( ) 0.198 P(C>5.5⏐C>4) =
Mean rate of damaging earthquake = 0.2x1/500 = 0.0004 P(building survives 100 years) = P(no damaging earthquake in 100 years) = e-0.0004x100 = e-0.04 = 0.96
(d)
P(survival of at least 4 structures during the earthquake)
⎛5⎞
⎛ 5⎞
= ⎜⎜ ⎟⎟(0.8) (0.2 ) + ⎜⎜ ⎟⎟(0.8) (0.2 ) ⎝ 5⎠ ⎝ 4⎠ = 0.4096+0.3277 = 0.737 4
1
5
0
mean rate earthquake that causes damage to at most 3 structures = 0.737x1/500 = 0.00147 P(at least four building will survive 100 years) = e-0.00147x100 = e-0.147 = 0.863
3.57 (a) Let A and B denote the pressure at nodes A and B, respectively. Since A is log-normal with mean = 10 and c.o.v. = 0.2 (small), we have ζ A ≅ 0.2, and λA = ln(µA) –
ζ A2 2
≅ ln(10) – 0.02 = 2.282585093
⇒ P(satisfactory performance at node A) ⎛ ln14 − 2.283 ⎞ − Φ ⎛ ln 6 − 2.283 ⎞ = Φ 1.788 − Φ −2.4560 = 0.955 = P (6 < A < 14) = Φ ⎜ ( ) ( ) ⎟ ⎜ ⎟ 0.2 0.2 ⎝ ⎠ ⎝ ⎠
(b) Let Ni denote the event of pressure at node i being within normal range; i = A,B. Given: P(NB) = 0.9 ⇒ P( N B ) = 0.1; P( N B | N A ) = 2×0.1 = 0.2 Hence P(unsatisfactory water services to the city) = P( N A ⋅ N B ) = P( N B | N A )P( N A ) = 0.2×(1 – answer to (a)) = 0.2×(1 – 0.955319) ≅ 0.0089 (c) The options are: (I) to change the c.o.v. of A to 0.15: repeating similar calculations as done in (a), we get the new values of: λA
ζA
2.2913
0.15
lower limit for Z –3.33
upper limit for Z 2.318
P(normal pressure at A) 0.98934
hence the probability of unsatisfactory water services, P( N B | N A )P( N A ) becomes 0.2×(1 – 0.9893) ≅ 0.0021 (II) to change P( N B ) to 0.05, and hence P( N B | N A ) = 2×0.05 = 0.10, thus P( N B | N A )P( N A ) = 0.10×(1 – 0.9556) ≅ 0.0044 ⇒ Option I is better since it offers a lower probability of unsatisfactory water services than II.
3.58 (a) fX(x) is obtained by “integrating out” the independence on y, 1
y3 ⎤ 6 6⎡ ( x + y 2 )dy = ⎢ xy + ⎥ 5 5⎣ 3 ⎦0 0 2 (0 < x < 1) = (3x + 1) 5 1
∴fX(x) =
(b) fY|X(y|x) =
f X ,Y ( x, y )
=
f X ( x)
∫
(6 / 5)( x + y 2 ) x + y2 =3 (2 / 5)(3 x + 1) 3x + 1 1
∫f
Hence P(Y > 0.5 | X = 0.5) =
Y |0.5 ( y |
x = 0.5)dy
0.5
1
⎡ y3 ⎤ 0 .5 + y 2 =3 dy = (3/2.5) ⎢0.5 y + ⎥ 1 .5 + 1 3 ⎦ 0.5 ⎣ 0.5 1
∫
= 0.65 1 1
(c) E(XY) =
∫∫
0 0 1
=
xyf X ,Y ( x, y ) dxdy =
6 5
1 1
∫ ∫ (x
2
y + xy 3 )dxdy
0 0
1
2 3 3 ydy + y dy = 1/5 + 3/20 = 7/20 = 0.35 50 50
∫
∫
⇒ Cov(X,Y) = E(XY) – E(X)E(Y) = 0.35 – (3/5)(3/5) = -0.01, while σX = {E(X2) – [E(X)]2}1/2 = [(13/30) – (3/5)2]1/2 = 0.271 σY = {E(Y2) – [E(Y)]2}1/2 = [(11/25) – (3/5)2]1/2 = 0.283, Hence the correlation coefficient, ρXY =
Cov( X , Y )
σ XσY
=
−0.01 0.271 × 0.283
≅ - 0.131
Summing over the last row, 2nd & 3rd columns of the given joint PMF table, we obtain P(X ≥ 2 and Y > 20) = 0.1 + 0.1 = 0.2
Given that X = 2, the new, reduced sample space corresponds to only the second column, where probabilities sum to (0.15 + 0.25 + 0.10) = 0.5, not one, so all those probabilities should now be divided by 0.5. Hence P(Y ≥ 20 | X = 2) = 0.25 / 0.5 + 0.1 / 0.5 = 0.35 / 0.5 = 0.7
f X and Y are s.i., then (say) P(Y ≥ 20) should be the same as P(Y ≥ 20 | X = 2) = 0.7. However, P(Y ≥ 20) = 0.10 + 0.25 + 0.25 + 0.0 + 0.10 + 0.10 = 0.80 ≠ 0.7,
hence X and Y are not s.i.
Summing over each row, we obtain the (unconditional) probabilities P(Y = 10) = 0.20, P(Y = 20) = 0.60, P(Y = 30) = 0.20, hence the marginal PMF of runoff Y is as follows:
fY(y) 0.6
0.2
0.2
10
20
30
y
Given that X = 2, we use the probabilities in the X = 2 column, each multiplied by 2 so that their sum is unity. Hence we have P(Y = 10 | X = 2) = 0.15×2 = 0.30, P(Y = 20 | X = 2) = 0.25×2 = 0.50, P(Y = 30 | X = 2) = 0.10×2 = 0.20, and hence the PMF plot:
fY|X(y|2) 0.5 0.3 0.2
By summing over each column, we obtain the marginal PMF of X as P(X = 1) = 0.15, P(X = 2) = 0.5, P(X = 3) = 0.35. With these, and results from part (d), we calculate E(X) = 0.15×1 + 0.5×2 + 0.35×3 = 2.2, 2 2 2 Var(X) = 0.15×(1 – 2.2) + 0.5×(2 – 2.2) + 0.35×(3 – 2.2) = 0.46, similarly E(Y) = 0.2×10 + 0.6×20 + 0.2×30 = 20, 2 2 2 Var(Y) = 0.2×(10 – 20) + 0.6×(20 – 20) + 0.2×(30 – 20) = 40, Also, E(XY) =
∑
x⋅y⋅f(x,y)
all
= 1×10×0.05 + 2×10×0.15 + 1×20×0.10 + 2×20×0.25 + 3×20×0.25 + 2×30×0.10 + 3×30×0.10 = 45.5 Hence the correlation coefficient is E ( XY ) − E ( X ) E (Y ) ρ= Var ( X ) Var (Y )
=
45.5 − 2.2 × 20
0.46 × 40 ≅ 0.35
4.1 Since it leads to an important result, let us first do it for the general case where X ~ N(λ, ζ), i.e.
f X ( x) =
1
1 ⎛ x −λ ⎞ ⎟ − ⎜⎜ 2 ⎝ ζ ⎟⎠
2
e 2π ζ To obtain PDF fY(y) for Y = eX, one may apply Ang & Tang (4.6), fY(y) = fX(g-1)
dg −1 dy
where y = g(x) = ex; ⇒ x = g-1(y) = ln(y) dg −1 d 1 ⇒ = ln y = dy dy y 1 since y = eX > 0 y Hence, expressed as a function of y, the PDF fY(y) is
=
fX(g-1)
=
1
dg −1 1 = fX(ln(y)) dy y 1 ⎛ ln y − λ ⎞ − ⎜⎜ ⎟⎟ 2⎝ ζ ⎠
2
e (non-negative y only) 2π yζ which, by comparison to Ang & Tang (3.29), is exactly what is called a log-normal distribution with parameters λ and ζ, i.e. if X ~ N(λ, ζ), and Y = eX, then Y ~ LN(λ, ζ).
Hence, for the particular case where X ~ N(2, 0.4), Y is LN with parameters λ being 2 and ζ being 0.4.
4.2 To have a better physical feel in terms of probability (rather than probability density), let’s work with the CDF (which we can later differentiate to get the PDF) of Y: since Y cannot be negative, we know that P(Y < 0) = 0, hence when y < 0: FY(y) = 0 ⇒ fY(y) = [FY(y)]’ = 0 But when y ≥ 0, FY(y) = P(Y ≤ y) = P(
1 mX 2 ≤ y ) 2
2y 2y ) ≤X≤ m m
= P( −
⎛ 2y ⎞ ⎛ 2y ⎞ ⎟ − FX ⎜ − ⎟ ⎜ m ⎟ ⎜ ⎟ m ⎝ ⎠ ⎝ ⎠
= FX ⎜
⎛ 2y ⎞ ⎟ ⎜ m ⎟−0 ⎝ ⎠
= FX ⎜
⎛ 2y ⎞ ⎟ ⎜ m ⎟ ⎝ ⎠
= FX ⎜ Hence the PDF, fY(y) =
d [FY(y)] dy
⎛ 2y ⎞ d 2y ⎟ ⎜ m ⎟ dy m ⎝ ⎠ 8y 2y 1 exp(− ) = 2 3 ma ma π 2my = fX ⎜
=
4 a3
2y 2y exp(− ) 3 πm ma 2
Hence the answer is
⎧4 ⎪ f Y ( y) = ⎨ a 3
2y 2y exp(− ) 3 πm ma 2
y≥0
4.3 A = volume of air traffic C = event of overcrowded (a)
T = total power supply = N ( µ T , σ T )
µ T = 100 + 200 + 400
Where
σ T = 15 2 + 40 2 + 40 2 = 58.5 (b)
P(Normal weather) = P(W) = 2/3 P(Extreme weather) = P(E) = 1/3 P(Power shortage) = P(S) = P(S⏐W)P(W)+P(S⏐E)P(E) = P(T<400)x2/3 + P(T<600)x1/3
400 − 400 2 600 − 400 1 )] × + [Φ ( )] × 58.5 3 58.5 3 2 1 = [Φ (0)] × + [Φ (3.42)] × 3 3
= [Φ (
= 0.5x0.667 + 0.99968x0.333 = 0.667
P( S W ) P(W )
=
0.5 × 0.667 = 0.5 0.667
(c)
P(W⏐S) =
(d)
P(all individual power source can meet respective demand) = P(N>0.15x400)P(F>0.3x400)P(H>0.55x400)
P( S )
220 − 400 120 − 200 60 − 100 )] )][1 − Φ( )][1 − Φ ( 40 40 15 = Φ (2.67)x Φ (2)x Φ (4.5)
= [1 − Φ (
= 0.09962x0.977x1 = 0.973 P(at least one source not able to supply respective allocation) = 0.027
4.4
(a) Let TJ be John’s travel time in (hours); TJ = T3 + T4 with
µ T = 5 + 4 = 9 (hours), and σ T = [32 + 12 + (2)(0.8)(3)(1)]1/2 = 3.847 (hours) J
J
Hence P(TJ > 10 hours) = 1 – P(
T J − µ TJ
σT
J
≤
10 − 9 ) 3.847
= 1 – Φ( 0.26) = 1 – 0.603 ≅ 0.397
(b) Let TB be Bob’s travel time in (hours); TB = T1 + T2 with
µT = 6 + 4 = 10 (hours), and B
σ T = [22 + 12]1/2 = 5 (hours) B
Hence P(TJ – TB > 1) = P(TB – TJ + 1 < 0), now let R ≡ TB – TJ + 1; R is normal with µR = µTB – µTJ + 1 = 10 – 9 + 1 = 2, σR = [ σ TB 2 + σ TJ 2]1/2 = (5 + 14.8)1/2 = 19.8 , hence P(R < 0) = Φ (
0−2
) 19.8 = Φ(-0.449) ≅ 0.327
(c) Since the lower route (A-C-D) has a smaller expected travel time of µ TJ = 9 hours as compared to the upper (with expected travel time = µTB = 10 hours), one should take the lower route to minimize expected travel time from A to D.
4.5 (a) To calculate probability, we first need to have the PDF of S. As a linear combination of three normal variables, S itself is normal, with parameters µS = 0.3×5 + 0.2×8 + 0.1×7 = 3.8 (cm) σS = 0.3 212 + 0.2 2 2 2 + 0.1212 ≅ 0.51 (cm) Hence
4 − 3 .8 ) 0.51 = 1 – Φ(0.3922) = 1 – 0.6526 ≅ 0.347
P(S > 4cm) = 1 –Φ (
(b) Now that we have a constraint A + B + C = 20m, these variables are no longer all independent, for example, we have C = 20 – A – B Hence S = 0.3 A + 0.2 B + 0.1(20 – A – B) ⇒ S = 0.2 A + 0.1 B + 2, with ρAB = 0.5. Thus µS = 0.2×5 + 0.1×8 + 2 = 3.8 (cm) as before, and
0.2 212 + 0.12 2 2 + 2 × 0.5 × 0.1 × 1 × 2 = 0.12 ≅ 0.346 (cm)
σS = Hence
P(S > 4cm) = 1 – Φ(
4 − 3.8
) 0.12 = 1 – Φ(0.577) ≅ 0.282
4.6 Q = 4A + B + 2C = 4A + B +2(30 – A – B) = 2A –B + 60, hence µQ = 2×5 – 8 + 60 = 62, 2 2 1/2 σQ = [4×3 + 1×2 + 2(2)(-1)(-0.5)(3)(2)] = (36 + 4 + 12)1/2 = 52 Q − µ Q 40 − 62 ) Hence P(Q < 40) = P( < σQ 52 = Φ(-3.0508) ≅ 0.00114
4.7 (a) Let Q1 and Q2 be the annual maximum flood peak in rivers 1 and 2, respectively. We have Q1 ~ N(35, 10), Q2 ~ N(25, 10) The annual max. peak discharge passing through the city, Q, is the sum of them, Q = Q1 + Q2, hence µ Q = µ Q1 + µ Q2 = 35 + 25 = 60 (m3/sec), and
σ Q2 = σ Q21 + σ Q2 2 + 2 ρ Q1Q2 σ Q1 σ Q2 2
2
= 10 + 10 + 2×0.5×10×10 = 300 (m3/sec), ⇒ σ Q = 300 ≅ 17.32 (m3/sec)
(b) The annual risk of flooding, p = P(Q > 100) = 1 – Φ(
100 − 60 ) 300
= 1 – Φ(2.309) = 1 – 0.9895 = 0.0105 (probability each year) Hence the return period is τ =
=
1 p
1 = 95.59643882 0.0105
≅ 95 years.
(c) Since the yearly risk of flooding is p = 0.0105, and we have a course of n = 10 years, we adopt a binomial model for X, the total number of flood years over a 10-year period. P(city experiences (any) flooding) = 1 – P(city experiences no flooding at all) = 1 – P(X = 0) = 1 – (1 – p)10 = 1 – (1 – 0.0105)10 = 1 – 0.989510 ≅ 10% (d) The requirement on p is, using the flooding probability expression from part (c): 1 – (1 – p)10 = 0.1 ÷ 2 = 0.05 ⇒ p = 1 – (1 – 0.05)1/10 = 0.0051, which translates into a condition on the design channel capacity Q0, following what’s done in (b), Q − 60 ) = 0.0051 1 – Φ( 0 300 Q − 60 ) = 0.9949 ⇒ Φ( 0 300 ⇒ Q0 = 60 + Φ-1(0.9949) = 60 + 2.57 300 = 104.5
300
∴Extending the channel capacity to about 104.5 m3/sec will cut the risk by half.
4.8
P(T<30)=P(T<30|N=0)P(N=0)+P(T<30|N=1)P(N=1)+P(T<30|N=2)P(N=2) 30 − 35 30 − 40 30 − 30 = Φ( ) × 0 .2 + Φ ( ) × 0 .5 + Φ ( ) × 0 .3 5 25 + 3 2 25 + 2 × 3 2 = 0.5 × 0.2 + Φ (−0.857) × 0.5 + Φ (−1.525) × 0.3 =0.217
4.9 N = total time for one round trip in normal traffic = N(30+20+40, = N(90, 15.52)
9 2 + 4 2 + 12 2 )
R = total round trip time in rush hour traffic = N(30+30+40, 9 2 + 6 2 + 12 2 ) = N(100, 16.16) (a)
P(on schedule in normal traffic) = P(N<20) = Φ(
(b)
120 − 90 ) = Φ (1.933) = 0.974 15.52
TA = Normal time taken for passenger starting from A = T2 + T3
4 2 + 12 2 ) = N(60, 12.65) 60 − 60 ) = Φ (0) = 0.5 P(TA <60) = Φ ( 12.65 = N(20+40,
(c)
Under rush hour traffic TA = N(30+40, TB = N(40, 12)
6 2 + 12 2 ) = N(70, 13.4)
60 − 70 ) = Φ (−0.746) = 0.227 13.4 60 − 40 P(TB<60) = Φ ( ) = Φ (1.667) = 0.952 12
P(TA<60) = Φ (
Percentage of passengers arriving in less than an hour = 0.227x1/3 + 0.952x2/3 = 0.7 (d)
P(45 < TB < 60) P(TB > 45) 60 − 40 45 − 40 Φ( ) − Φ( ) Φ (1.667) − Φ (0.417) 0.952 − 0.661 12 12 = = = = 0.858 45 − 40 1 − Φ (0.417) 0.339 1 − Φ( ) 12 P(TB <60⏐ TB >45) =
4.10 (a) D=S1-S2
µD = µS − µS = 2 − 2 = 0 1
2
Var ( D) = Var ( S1 ) + Var ( S 2 ) − 2 ρ Var ( S1 )Var ( S 2 ) = ( 0 .3 × 2 ) 2 + ( 0 .3 × 2 ) 2 − 2 × 0 .7 ( 0 .3 × 2 ) 2 = .216 (b) P (−.5 < D < .5) = Φ (
0.5 − 0 0.216
) − Φ(
− 0.5 − 0 0.216
) = Φ (1.076) − Φ (−1.076) = 0.718
4.11 (a) X 5 = X 1 + X 2 + X 3
µ 5 = µ1 + µ 2 + µ 3 = 10 + 15 + 20 = 45 σ 52 = σ 12 + σ 22 + 2 ρ1, 2σ 1σ 2 = 3 2 + 3 2 + 2 × 0.6 × 3 × 3 = 28.8 ∴ σ 5 = 5.37 (b)
V1 = 60 X 1 V2 = 60 X 2 ∴ P(V2 − V1 > 400) = P( Z > 400) ∴ µ Z = 60 µ 2 − 60 µ1 = 60 × 15 − 60 × 10 = 300
σ Z2 = 60 2 σ 22 + 60 2 σ 12 − 2 ρ 1, 2 × 60 2 σ 1σ 2 = 25920 ∴ σ Z = 161 ∴ P(V2 − V1 > 400) = 1 − Φ ( (c) X 5 = X 1 + X 2 + X 3
400 − 300 ) = 1 − Φ (.621) = .267 161
µ 5 = µ1 + µ 2 + µ 3 = 10 + 15 + 20 + 3n = 45 + 3n σ 52 = σ 12 + σ 22 + 2 ρ1, 2σ 1σ 2 = 3 2 + 3 2 + 2 × 0.6 × 3 × 3 = 28.8 ∴ σ 5 = 5.37 ∵ P( X 5 > 70) = .05 ∴ P( X 5 < 70) = .95 ⎛ 70 − 45 − 3n ⎞ Φ⎜ ⎟ = .95 5.37 ⎝ ⎠ ⇒ 25 − 3n = 5.37Φ −1 (.95) = 5.37 *1.645 = 8.839 ⇒ n = 5 .4
4.12
(a)
P(X>20∪Y>20)=P(X>20)+P(Y>20)-P(X>20)P(Y>20)
= (1 − Φ (
(b)
20 − 20 20 − 15 20 − 20 20 − )) + (1 − Φ ( )) − (1 − Φ ( )) × (1 − Φ ( 4 3 4 3
=0.5+0.0478-0.5 × 0.0478 =0.524 Z in normal distribution with µ Z = 0.6 × 20 + 0.4 × 15 = 18
σ Z = 0.6 2 4 2 + 0.4 2 3 2 = 2.683 20 − 18 ) = 1 − 0.7718 = 0.2282 2.683 µ Z = 0.6 × 20 + 0.4 ×15 = 18
∴ P( Z > 20) = 1 − Φ ( (c)
σ Z = 0.6 2 4 2 + 0.4 2 3 2 + 2 × 0.8 × 0.6 × 0.4 × 4 × 3 = 3.436 ∴ P( Z > 20) = 1 − Φ (
20 − 18 ) = 1 − 0.72 = 0.28 3.436
4.13
(a)
(b)
(c)
P(X≥2)=1-P(X=0)-P(X=1) ⎛5⎞ ⎛5⎞ = 1 − ⎜⎜ ⎟⎟0.6 0 0.4 5 − ⎜⎜ ⎟⎟0.610.4 4 ⎝0⎠ ⎝1⎠ = 1 − 0.4 5 − 5 × 0.6 × 0.4 4 = 0.046 NH, NB are the number of highway and building jobs won P(NH =1∩ NB =0)=P(NH =1)P(NB =0) ⎛ 3⎞ ⎛ 2⎞ = ⎜⎜ ⎟⎟0.610.4 2 ⎜⎜ ⎟⎟0.6 0 0.4 2 ⎝1⎠ ⎝0⎠ =0.046 T=H1+H2+B where H1, H2 and B are the profits from the respective jobs. T in Normal distribution with µ T = 100 + 100 + 80 = 280
σ T = 40 2 + 40 2 + 20 2 = 60
300 − 280 ) = 1 − Φ (0.333) = 0.2695 60 µ T = 100 + 100 + 80 = 280
∴ P(T > 300) = 1 − Φ ( (d)
σ T = 40 2 + 40 2 + 20 2 + 2 × 0.8 × 40 × 40 = 78.5 ∴ P(T > 300) = 1 − Φ (
300 − 280 ) = 1 − Φ (0.255) = 0.4 78.5
4.14
S: the amount of water available. S follows LogNormal Distribution with E(S)=1, c.o.v=0.4 ζ S2 = ln(1 + 0.4 2 ) = 0.14842
λ S = ln µ S − 12 ζ S2 = ln 1 − 0.5 × 0.14842 = −0.07421 D: the total demand of water. E(D)=1.5, c.o.v=0.1 ζ D2 = 0.12 = 0.01 λ D = ln µ D − 12 ζ D2 = ln 1.5 − 0.5 × 0.01 = 0.4005 P(water shortage)=P(D>S) =P(D/S>1) =P((Z=ln(D/S))>0) Z follows Normal Distribution with: Var ( Z ) = ζ 2 D + ζ S2 = 0.12 + 0.148 = 0.158 µ Z = λ D − λ S = 0.4005 − (−0.07421) = 0.4747 0 − 0.4747 ⇒ P ( Z > 0) = 1 − Φ ( ) = 0.883 0.158
4.15 (a)
P is lognormally distributed with λP = λC + λR + 2λV – ln2 = (ln1.8 – 0.5x0.22) + (ln2.3x10-3 – 0.5x0.12) + 2[ln120-0.5xln(1+0.452)] – ln2 = 0.568 + (-6.08) + 2(4.6953) – 0.693 = 3.186 ξP
ξ + ξ + 4ξ = 0.2 + 0.1 + 4 × ln(1 + 0.45 ) = 0.2 + 0.1 + 4 × 0.1844 2 C
2 R
2 V
2
2
2
2
= 0.887
ln 30 − 3.186 ) = 1 − Φ (0.243) = 0.596 0.887
(b)
P(P>30) = 1 − Φ(
(c)
C is lognormally distributed with λC = ln90 – 0.5x0.152 = 4.4886 ξC = 0.15 P(Failure of antenna) = P(C
ξ C2 + ξ P2 = 0.15 2 + 0.887 2 = 0.9
Hence, P(Failure of antenna) = P(B<1) = Φ(
(d)
ln 1 − λ B
ξB
) = Φ(
− 1 .7 ) = Φ (−1.89) = 0.029 0 .9
Mean rate of wind storm causing failure = 1/5 x 0.029 = 0.0058 P(antenna failure in 25 years) = 1-P(no damaging storm in 25 years) = 1-e-0.0058x25 = 0.135
(e)
P(at least two out of 5 antenna failures) = 1 – P(X=0) – P(X=1) = 1 – 0.8655 – 5(0.135)(0.865)4 = 0.138
2
=
4.16 (a) P(failure)=P(L>C)=P(C/L<1)=P(Z<1) Z in LN with λ Z = λC − λ L
ζ Z = ζ C2 + ζ L2 in which ζ C = 0.2
1 2
λC = ln 20 − (0.2) 2 = 2.259
ζ L = ln(1 + δ 2 ) = .294 λ L = ln 10 − 12 (0.294) 2 = 2.259 ∴ λ Z = 2.976 − 2.259 = 0.717
ζ Z = 0.2 2 + 0.294 2 ∴ PF = Φ (
= 0.356
ln 1 − 0.717 ) = Φ (−2.014) = 1 − 0.978 = 0.022 0.356
(b) T=C1+C2 E(T)=E(C1)+E(C2)=20+20=40
Var (T ) = Var (C1 ) + Var (C 2 ) + 2 ρσ C1 σ C2 = 57.6
δT = (c)
57.6 = 0.19 40
some other distribution
4.17
(a)
C=F+B µ C = µ F + µ B = 20 + 30 = 50
σ C = (0.2 × 20) 2 + (0.3 × 30) 2 (b)
= 9.85
T=C1+C2=2 µ C =0.197
σ T = 9.85 2 + 9.85 2 + 2 × 0.8 × 9.85 2 = 18.69 ∴ δ T = 18.69 (c)
= 0.187 100 P(failure)=P(T
σZ
= Φ(
− (100 − 50) 18.69 2 + (.3 × 50) 2
− 50 ) 23.95 = 0.0184 = Φ(
4.18 16
(a)
F = 18 + ∑ Ai i =1
µ F = 18 + 16 × 0.1 = 19.6
σ F = (0.3 × 0.1) 2 × 16 = 0.12 P(F>20)=1- Φ ( (b) i.
20 − 19.6 ) = 0.00043 0.12
P(no collapse)=P(C’)=P(F
σ Z = (0.12) 2 + (0.01 × 20) 2
= 0.233
0 − (−0.4) ) = 0.957 0.233 F=18+16A µ F = 18 + 16 × 0.1 = 19.6 P(C’)= Φ (
ii.
σ F = (0.3 × 0.1) 2 × 16 2 = 0.48 σ Z = (0.48) 2 + (0.01 × 20) 2 P(C’)= Φ (
0 − (−0.4) ) = 0.779 0.52
= 0.52
4.19 (a)
a is lognormal with mean 0.3g and c.o.v. of 25% W = 200 kips F = wa/g is also lognormal with λF = ln 200 + λa = 5.298 – 1.204 = 4.094 ξF = ξa = 0.25 R = frictional resistance = WC Where C = coefficient of friction is lognormal with median 0.4 and a c.o.v. of 0.2 Hence R is lognormal with λR = ln 200 + λC = ln200 + ln0.4 = 4.382 ξR = ξC = 0.2 P(failure) = P(R
(b)
− λR + λF
ξ R2 + ξ F2
) = Φ(
− 4.382 + 4.094 ) = Φ (−0.9) = 0.184 0.32
P(none out of five tanks will fail) = (0.184)5 = 0.00021
4.20 For any one car, its average waiting time is µT = 5 minutes, with standard deviation σT = 5 minutes. Let W be the total waiting time of 50 cars. W is the sum of 50 i.i.d. random variables, hence, according to CLT, W is approximately normal with mean µW = 50×µT = 50×5 min. = 250 min., and standard deviation σW = 50 ×σT = 50 ×5 min. Hence the probability P(W < 3.5 hrs) = P(W < 210 mins) W − µ W 210 − 250 < ) = P( σW 5 50 = P(Z < -1.13137085) ≅ 0.129
4.21 (a) First, let’s calculate the parameters of F and X: ξF = [ln(1 + δF2)]0.5 = [ln (1 + 0.2 )] = [ln(1.04)] 2 0.5 ∴ λF = ln µF – ξF /2 = ln(0.2 / 1.04 ), similarly 2 0.5 0.5 ξX = [ln(1 + δX2)]0.5 = [ln (1 + 0.3 )] = [ln(1.09)] 2 0.5 ∴λx = ln µx – ξx /2 = ln(10 / 1.09 ) 2 0.5
0.5
Now, let M denote the bending moment at A. Since M = FX, ln M = ln F + ln X (i.e. normal + normal), hence M is lognormal (because ln M is normal), with parameters 2 λM = λF + λX = ln( ), and 1.04 × 1.09 0.5 0.5 ξM = (ξF2 + ξX2) = [ln(1.04×1.09)] , hence P(M > 3) = P(
ln M − λ M
ξM
>
ln 3 − ln(2 / 1.04 × 1.09) ln(1.04 × 1.09)
)
= 1 – Φ(1.322063427) ≅ 0.093
(b) Let Mi be the moment at A due to the i-th force, where i = 1,2,…,50. Since each Mi is
identically distributed as M in part (a), the lognormal parameters of Mi are λ = 0.630447976 and ξ = 0.354116378. From these, we can calculate the mean and standard deviation of Mi as µi = exp(λ + ξ2/2) = 2, σi = µi [ exp(ξ2 ) – 1] 0.5 = 0.731026675 The total bending moment at A is T = M1 + M2 + … + M50, the sum of a large number of identically distributed, independent RVs, hence we may apply the central limit theorem: T ~ N(50×2, 50 ×0.731026675) T − µT 120 − 50 × 2 > ∴P(T > 120) = P( ) σT 50 × 0.731026675 = 1 – Φ( 3.869116163) = 1 – 0.999945364 ≅ 0.000055
4.22 (a) Let X be the monthly salary (in dollars) of a randomly chosen assistant engineer. X has a uniform distribution between 10000 and 20000, covering an area of 20000 − 16000 ×1 = 0.4 20000 − 10000
from x = 16000 to x = 20000. (b) Using properties of an uniform random variable, µX = (10000+20000)/2 = 15000, (20000 − 10000) 5000 σX = . = 12 3 Let Y be the mean monthly salary of 50 assistant engineers. By the central limit theorem, Y is approximately normally distributed with σ 5000 µY = µX = 15000 and σY = X = , 50 150 hence the desired probability is 16000 − 15000 ) P(Y > 16000) = 1 –Φ ( 5000 / 150 = 1 –Φ(2.45) ≅ 0.007
(c) An individual’s probability of exceeding $16000 is much higher than that for the mean of a group of people, as X has a much larger standard deviation than Y, though they have the same mean. Uncertainty is reduced as sample size increases, and “collective behavior” (i.e. average value) becomes very centralized (i.e. almost constantly at $15000) around the mean, while individual behavior can differ from the mean significantly.
4.23 (a) Let C denote car weight and T denote truck weight. The total vehicle weight (in kips) on the bridge is V = C1 + C2 + … +C100 + T1 + T2 + … + T30 Since V is a linear combination of independent random variables, it mean and variance can be found as E(V) = 100×E(C) + 30×E(T) = 100×5 + 30×20 = 1100 (kips); 2 2 Var(V) = 100×Var(C) + 30×Var(T) = 100×2 + 30×5 = 1150 2 (kip ), Hence the c.o.v. 0.5 δV = 1150 / 1100 ≅ 0.031 (b) Assuming that the distribution of V approaches normal due to CLT, P(V > 1200) =1 – Φ (
1200 − 1100 ) 1150
= 1 – Φ(2.948839) ≅ 0.0016 (c) (i)
(ii)
Let D denote the total dead load in kips, where D ~ N(1200, 120) The difference S = V – D is again normal, 2 1/2 0.5 S ~ N(1100 – 1200, (1150 + 120 ) ) = N( –100, 15550 ), hence P(V > D) = P(V – D > 0) = P(S > 0) S − µ S 0 − (−100) > ) = P( σS 15550 = 1 – Φ(0.80192694) ≅ 0.211
Let T denote the total (dead + vehicle) weight, T = V + D. T ~ N(1100 + 1200, 0.5 15550 ), hence P(V + D > 2500) = P(T > 2500) T − µ T 2500 − 2300 > = P( ) σT 15550 = 1 – Φ(1.60385388) ≅ 0.054
4.24 (a) Let B be the total weight of one batch. µB = 40×2.5 kg = 100 kg (b) Since n > 30, we may apply the Central Limit Theorem: B is approximately normal with µB = 100 kg and σB =
40 × 0.1 kg, hence
101 − 100
) 40 × 0.1) = 1 –Φ (1.581) = 1 – 0.9431 ≅ 5.69% 101 − 100 ) (c) In this case, P(penalty) = P(B > 101) = 1 –Φ ( 40 × 1) = 1 –Φ (0.158) = 1 – 0.562816481 ≅ 43.7% P(penalty) = P(B > 101) = 1–Φ (
This penalty probability is much higher, caused by the large standard deviation which makes the PDF occupy more area in the tails. Hence a large standard deviation (i.e. uncertainty) is undesirable.
4.25 Over the 50-year life, the expected number of change of occupancy is 25. (a)
The PDF of the lifetime maximum live load Yn or Y20 in this case is, from Eq. 3.57,
f Yn ( y ) = f Y20 ( y ) = 20[ Fx ( y )]19 f x ( y ) However X is lognormal with parameters λ, ξ whose values are ξ=
ln(1 + 0.3 2 ) = 0.294
λ = ln12 – 0.5x0.2942 = 2.44
1 ln y − 2.44 2 ) ] exp[− ( 0.294 2 2π (0.294) y ln y − 2.44 ) FX ( y ) = Φ( 0.294 ln y − 2.44 19 1 1 ln y − 2.44 2 ] exp[− ( f Y20 ( y ) = 20[Φ )] 0.294 0.737 y 2 0.294
(b)
1
f X ( y) =
Hence
By following the results of Examples 3.34 and 3.37, Y20(y) will converge asymptotically to the Type I distribution with parameters un = 2.94( 2 ln 20 − and α n =
ln ln 20 + ln 4π
)+2.44 = 2.187
2 2 ln 20
2 ln 20 = 8.326 0.294
νn = e2.187 = 8.9 k = 8.326
f Y20 ( y ) =
8.326 8.9 9.326 8.9 ( ) exp[−( ) 8.326 ] y y 8.9
4.26 (a)
f X ( x) =
ν (νx) k −1 e −νx Γ(k )
Then, x
FX ( x) = ∫
ν (νz ) k −1 e −νz Γ(k )
0
dz
The CDF of the largest value from a sample of size n is (Eq. 4.2): x
FYn ( y ) = [ FX ( y )] = [ ∫ n
ν (νz ) k −1 e −νz Γ(k )
0
dz ]n
whereas, the PDF is (Eq. 4.3): n −1
fYn ( y ) = n[ FX ( y )]
x
f x ( y ) = n[ ∫
ν (νz ) k −1 e −νz Γ(k )
0
(b)
n −1
dz ]
ν (νx) k −1 e −νx Γ(k )
If the distribution of the large value is Type I asymptotic, then according Eq. 4.77,
d
1
lim dx [ h ( x) ] = 0 x →∞
n
The hazard function hn(x) is (Eq. 4.28):
hn ( x) =
f X ( x) = 1 − FX ( x)
ν (νx) k −1 e −νx / Γ(k ) x
1 − ∫ν (νz ) k −1 e −νz dz / Γ(k ) 0
ν x k
=
k −1
x
e
Γ(k ) − ∫ ν z k
0
−νx
k −1
e
x k −1 e −νx
= −νz
dz
x
ν Γ(k ) − ∫ z k −1 e −νz dz −k
0
x
d
lim dx [ h x →∞
ν − k Γ(k ) − ∫ z k −1e −νz dz
1 d ] = lim [ x → ∞ dx n ( x)
0 k −1 −νx
x e
]
Using Leibnitz rule for the derivative of the integral, the above limit becomes, x
− x k −1e −νx x k −1e −νx − [ν − k Γ(k ) − ∫ z k −1e −νz dz ]e −νx [(k − 1) x k − 2 − νx k −1 0 k −1 −νx 2
lim
(x e
x →∞
)
x
= −1−
[ν − k Γ(k ) − ∫ z k −1e −νz dz ](k − 1 − νx) 0
lim x→∞
x k e −νx
Recall that ∞
∫z
k −1 −νz
e dz =
0
Γ(k )
νk x
−k k −1 −νz lim{[ν Γ(k ) − ∫ z e dz ](k − 1 − νx)} = 0 x →∞
0
and k −νx
lim ( x e
)=0
x →∞
Therefore, using the L’Hospital’s rule: x
− 1 − lim
− x k −1e −νx (k − 1 − νx) − ν [ν − k Γ(k ) − ∫ z k −1e −νz dz ]
x→∞
e −νx x k −1 (k − νx)
− ν [ν
−k
0
x
Γ(k ) − ∫ z k −1 e −νz dz ]
− ( k − 1 − νx ) 0 − lim −νx k −1 k − νx e x ( k − νx ) x →∞ x →∞ ∞ The second term is of the form . Again, using L’Hospital’s rule: ∞ − (k − 1 − vx) ν = = −1 lim k − νx −ν x →∞ 0 The third term is of the form , thus, 0 νx k −1 e −νx =0 lim −νx [−νx k −1 (k − νx) + (k − 1) x k − 2 (k − νx) − νx k −1 ] x →∞ e = −1−
lim
since the order of the polynomial in the denominator is higher than that of the numerator. Therefore, the distribution of the largest value converges to the Type I distribution.
4.27 (a)
Let Z = X-18; Z follows an exponential distribution with mean = 3.2, i.e. λ = 1/3.2 Following the result of Example 3.33, the largest value of an exponentially distributed random variable will approach the Type I distribution. Also, for large n, the CDF of the largest value
FYn ( y ) = exp(−ne − λy ) which can be compared with the Type I distribution
FYn ( y ) = exp(−e −α n ( y −un ) ) By setting ne-λy = e
−α n ( y − u n )
We obtain αn = λ and
=e
un =
−α n y α n u n
e
ln n
λ
For a 1 year period, the number of axle loads is 1355. Hence mean maximum axle load is
u Y1355 = 18 + [u n +
σY
1355
=
ln 1355 0.577 γ = + ] = 42.9 α n 1 / 3.2 1 / 3.2
π 2 π 2 (3.2) 2 = = 16.8 6 6α n2
∴δ Y1355 =
16.8 = 0.39 42.9
Similarly for period of 5, 10 and 25 years, the number of axle loads are 6775, 13550 and 33875 respectively. The corresponding mean values and c.o.v. are as follows: µY6775 = 48 σY6775 = 16.8 δY6775 = 0.35 (b)
µY13550 = 50.3 σY13550 = 16.8 δY13550 = 0.34
µY33875 = 53.2 σY33875 = 16.8 and δY33875 = 0.316
For a 20 year period, n = 27100
FYn ( y ) = exp(−e −α n ( y −18−un ) ) = exp[−e − (1/ 3.2 )( y −18−(ln 27100 )3.2 ) ] Hence the probability that it will subject to an axle load of over 80 tone is
1 − FY27100 (80) = 1 − exp[−e − (1/ 3.2)(80−18−(ln 27100 ) 3.2 ) ] = 1 – 0.999896 = 0.000104 (c)
For an exceedance probability of 10%, the “design axle load” L can be obtained from
1 − exp[−e − (1/ 3.2 )( L −18−(ln 27100) 3.2) ] ≡ 0.1 Hence, exp[−e − (1/ 3.2 )( L −50.7 ) ] = 0.9 L = 57.9 tons
4.28 (a)
Daily DO level = N(3, 0.5) From E 4.18, the largest value for an initial variate of N(µ , σ) follows a Type I Extreme Value distribution. Because of the symmetry between the left tails of the normal distribution, it can be shown that the monthly minimum DO level also follows a Type I smallest value distribution with
u1 = − 2 ln 30 +
ln ln 30 + ln 4π
2 2 ln 30 α 1 = 2 ln 30 / 0.5 = 5.21
+ 3 = 1.112
Similarly, for the annual minimum DO, it also follows the Type I smallest value distribution with
u1 = − 2 ln 365 +
ln ln 365 + ln 4π
+ 3 = −3.435 + 1.227 + 3 = 0.792
2 2 ln 365
α 1 = 2 ln 365 / 0.5 = 6.87 (b)
P[(Y1)30 < 0.5] = 1 – exp[-e5.21(0.5-1.112)] = 0.04 P[(Y1)365 < 0.5] = 1 – exp[-e6.87(0.5-0.792)] = 0.125
(c)
For Type I smallest value distribution
µ 1 = u1 −
0.577
α1
;
π2 σ1 = 2 6α 1
Hence for monthly minimum DO level, 0.577 = 1.0mg / l ; µ 1 = 1.112 − 5.21 Similarly for the annual minimum level, 0.577 = 0.708mg / l ; µ 1 = 0.792 − 6.87
σ1 =
π2 6 × 5.212
σ1 =
= 0.06mg / l
π2 6 × 6.87 2
= 0.035mg / l
4.29 Let Y be the maximum wind velocity during a hurricane Y follows a Type I asymptotic distribution with µY = 100 and σY = 0.45 From Equations 3.61a and 3.61b, the distribution parameters un and αn can be determined as follows:
π2 = (0.45 × 100) 2 or 2 6α n π αn = = 0.0285 6 (0.45 ×100) 0.5772 u n = 100 − = 79.75 0.0285
(a)
P(Damage during a hurricane) = P(Y>150) = 1 − FY (150) = 1 − exp[−e −0.0285(150 − 79.75) ] = 0.126
(b)
Let D be the revised design Hence, 1 − FY ( D ) = 1 − exp[−e −0.0285( D − 79.75) ] = 0.063 D = 175.6 kph
(c)
Mean rate of damaging hurricane = ν = (1/200) × 0.126 = 0.00063 P(Damage to original structure over 100 years) = 1 – P(no damaging hurricane in 100 years) = 1 – exp(-0.00063 × 100) = 0.061 P(damage to revised structure over 100 years) = 1 – e-(0.00063/2)(100) = 0.031
(d)
P(at least one out of 3 structure with original design will be damaged over 100 years) = 1 – P(none of the 3 structure damaged) = 1 – (1-0.061)3 = 0.172
4.30 (a)
Daily maximum wind velocity is N(40, 10) The maximum over 3 months will converge to a Type I largest value distribution whose CDF is (from Example 4.18
FYn ( y ) = exp[−e −α n ( y − u n ) ] with
un = 2 ln n −
ln ln n + ln 4π +µ 2 2 ln n
α n = 2 ln n / σ For a 3 month period, use n = 91 Hence
un = 2 ln 91 −
ln ln 91 + ln 4π + 40 = 42.1 2 2 ln 91
α n = 2 ln 91 / 10 = 0.3 (b)
The most probable wind velocity during the 3-month period is 42.1 mph P(Damage) = P(Yn > 70) = 1 − exp[−e −0.3( 70 − 42.1) ] = 0.00023
(c)
Mean of Yn = un + γ/αn = 42.1 + 0.577/0.3 = 44 mph σ of Yn = π /( 6α n ) = π /(0.3 6 ) = 4.277 and c.o.v. of Yn = 4.277/44 = 0.097
4.31 (a)
Daily maximum wind velocity is lognormal with mean 40 and c.o.v. of 25%. The maximum wind velocity over a 3-month period will converge to a Type II, or the Fisher-Tippett distribution, whose CDF is
νn
FYn ( y ) = exp[−(
y
)k ]
with the most probable value νn and shape parameters k
ν n = eu and
k = αn ln ln n + ln 4π un = ζ ( 2 ln n − )+λ 2 2 ln n
αn =
and
n
2 ln n
ζ
using the results obtained in Example 4.21 In this case, n = 91 ; ζ = 0.25
1 2
λ = ln 40 − (0.25) 2 = 3.658 Hence the most probable wind velocity during erection period is
ν n = exp[0.25( 2 ln 91 − = exp[4.18] = 65.4 mph Moreover,
k = 2 ln 91 / 0.25 = 12 (b)
P(Damage) = P(Yn > 70) = 1 − exp[−( = 0.36
65.4 12 ) ] 70
ln ln 91 + ln 4π ) + 3.658 2 2 ln 91
4.32
h=
LV 2 f 2 Dg
Variable L D f V (a)
Mean Value 100 ft 1 ft 0.02 9.0 fps
c.o.v. 0.05 0.15 0.25 0.2
std. Deviation 5 ft 0.15 ft 0.005 1.8 fps
Assuming the above random variables are statistically independent, the first order mean and variance of the hydraulic head loss h are, respectively,
µh ≈
100(9) 2 (0.02) = 2.516 ft 2(1)(32.2)
µL µ 2 2 µL µ 2 2 µ 2µ µ 2 µV2 µ f 2 2 2 ) +σD( 2 ) +σ f ( ) + σ V2 ( L ) σ ≈σ ( µ D (2 g ) µ (2 g ) µ (2 g ) µ (2 g ) 2 h
V
2 L
V
D
D
V
f
D
=
100 × 2 × 9 × 0.02 2 100 × 92 2 100 × 92 2 92 × 0.02 2 ) ) + 1 .8 2 ( ) + 0.0052 ( ) + 0.152 ( 2 52 ( 1(2 × 32.2) 1(2 × 32.2) 1 (2 × 32.2) 1(2 × 32.2) = 0.0158+355.9433+0.3955+1.01246 = 357.367 Hence, σ h ≈ 18.9 ft (b)
The second order mean of the hydraulic head loss
1 2
µ h = 2.516 + σ V2 (
µL µV µ f 1 (2)) ) + σ D2 ( 3 µ D (2 g ) 2 µ D (2 g ) 2µ L µ f
2
2 1 2 2 × 100 × 0.02 1 2 100 × 9 × 0.02 ) + 0.15 ( 3 (2)) = 2.516 + 1.8 ( 2 1(2 × 32.2) 2 1 (2 × 32.2)
= 2.516 + 0.1006 + 0.056 = 2.673 ft
4.33
D=
PL3 , 3EI
Variable P E b h
I=
bh 3 12
⇒D=
Mean Value 500 lb 3,000,000 psi 72 inch 144 inch
4 PL3 Ebh 3 c.o.v. 0.2 0.25 0.05 0.05
std. Deviation 100 lb 750,000 psi 3.6 inch 7.2 inch
L = 15 ft = 180 inch (a)
Assuming the above random variables are statistically independent, the first order mean and variance of the deflection D are, respectively,
4 × 500 × 1803 = 1.8 × 10 −5 inch 3 3000000 × 72 × 144 4 L3 2 4 µ P L3 2 4 µ P L3 × 3 2 4 µ P L3 2 2 2 2 2 2 ) +σE( 2 ) + σb ( ) +σh ( ) σD ≈ σP( 3 3 2 3 4
µD ≈
µ E µb µ h
+ 2 × ρ bh × σ b × σ h × = 100 2 (
µ µb µ h E
4µ P L3
× 3
µ E µ b2 µ h
µ E µb µ h
µ E µb µ h
4µ P L3 × 3
µ E µ b µ h4
4 × 500 × 180 3 2 4 × 500 × 180 3 2 4 × 180 3 2 2 2 ) ) 3 . 6 ( ) 750000 ( + + 3e6 × 72 2 × 144 3 (3e6) 2 × 72 × 144 3 3e6 × 72 × 144 3
+ 7.2 2 (
4 × 500 × 180 3 × 3 2 4 × 500 ×180 3 × 3 4 × 500 × 180 3 × 3 ) + 2 × 0 . 8 × 3 . 6 × 7 . 2 × × 3e6 × 72 ×144 4 3e6 × 72 2 ×144 3 3e6 × 72 ×144 4
= 1.308x10-11 + 2.044x10-11 + 8.176x10-13 + 7.359x10-12 + 3.925x10-12 = 4.562x10-11 Hence, σ D ≈ 6.75x10-6 ft (b)
The second order mean of the deflection
3 4 µ P L3 (−3)(− 4 µ P L3 (−1)(−2) 2 1 2 4 µ P L (−1)(−2) 2 2 µ D = 1.8 × 10 + [σ E ( ) +σb ( ) +σ h ( 2 µ 3E µ b µ h3 µ E µ b3 µ h3 µ E µ b µ h5
−5
=
1.8 × 10 −5 +
750000 2 × 2 3 .6 2 × 2 7.2 2 × 12 4 × 500 × 180 3 + + ] [ (3e6) 3 × 72 × 144 3 3e6 × 72 3 × 144 3 3e6 × 72 × 144 5 2
= 1.94x10-5 ft
4.34 (a) Using Ang & Tang Table 5.1, E(X) = (4 + 2) / 2 = 3, Var(X) = (4 - 2)2 / 12 = 1/3; E(Y) = 0 + (3 - 0) / (1 + 2) = 1, Var(Y) = 1×2×(3 - 0)2/(1+2)2(1+2+1) = 1/2; Also, for W, since we know its median is 1, i.e. 1
∫ λe λ dx = 0.5 ⇒ 1 − e λ = 0.5 ⇒ λ = ln 2 , hence − x
−
0
E(W) = 1/λ ≈ 1.443, Var(W) = 1/λ2 ≈ 2.081 (b) Since Z = XY2W0.5, we compute the approximate values E(Z) ≈ E(X) [E(Y)]2 [E(W)]0.5 = 3×12×1.4430.5 ≈ 3.6, and
⎡ ∂Z ⎤ ⎡ ∂Z ⎤ ⎡ ∂Z ⎤ Var(Z) ≈ ⎢ Var ( X ) + ⎢ ⎥ Var (Y ) + ⎢ Var (W ) ⎥ ⎣ ∂X ⎦ µ ⎣ ∂Y ⎦ µ ⎣ ∂W ⎥⎦ µ 2
[
2
]
2
[
2
]
2
[
]
2
= Y 2W 0.5 µ Var ( X ) + 2 XYW 0.5 µ Var (Y ) + 0.5 XY 2W −0.5 µ Var (W ) where µ denotes "evaluated at the mean values X = 3, Y = 1, W = 1.443", hence Var(Z) ≈ 29.7, 29.7 ≈ 1.51 Thus the c.o.v. of Z is 3.6
4.35 (a) Let X’ ≡ ln(X). Since X is log-normal, X’ is normal with parameters σX’ ≅ δX = 0.20, and µX’ ≅ ln 100 – δX2 / 2 = ln 100 – 0.202 / 2 = 4.585170186 When L is a constant (0.95), we have Y = (1 – 0.95)X = 0.05 X Taking the log of both sides and letting Y’ ≡ ln(Y) ⇒ Y’ = ln 0.05 + X’, i.e. Y’ is a constant plus a normal variate, hence Y’ is normal with parameters µY’ = ln 0.05 + µX’ ≅1.589437912, and σY’ = σX’ ≅ 0.20 Hence P(acceptable) = P(Y ≤ 8) = P(Y’ ≤ ln 8) Y '− µ Y ' ln 8 − 1.589437912 = P( ) ≤ σY' 0.20 = Φ(2.450018146) ≅ 0.993 (b) (i)
When L is also a random variable, Y = (1 – L)X is a non-linear function of two random variables, hence we need to estimate its mean and variance by E(Y) ≅ Y(L = µL, X = µX) = (1 – µL)(µX) = (1 – 0.96)(100) = 4, and
⎛ ∂Y ⎞ ⎛ ∂Y ⎞ Var(Y) ≅ ⎜ ⎟ Var ( L) + ⎜ ⎟ Var ( X ) ⎝ ∂L ⎠ µ ⎝ ∂X ⎠ µ 2
2
=
(− X )2X =100,L =0.96 (0.012 ) + (1 − L )2X =100,L =0.96 (100 × 0.2) 2
= 10000×0.0001 + 0.0016×400 = 1.64
(ii)
Assuming a log-normal distribution for Y and using its estimated mean and variance from (i), we have the approximate parameter values δY ≅
1.64
, hence
ξY =
ln(1 + δ Y2 )
≅
ln(1 + 1.64 / 16) = 0.312; 2 λY ≅ ln 4 – 0.312 / 2 = 1.338, thus ln 8 − 1.338 P(Y ≤ 8) = Φ ( ) 0.312 = Φ(2.375) ≅ 0.991 No, this offer from the contractor does not yield a larger performance acceptance probability.
4.36 (a) E(Qc)=0.463E(n)-1E(D)2.67E(S)0.5=0.463*0.015-1*3.02.67*0.0050.5=41.0ft3/sec Var(Qc)= [
=
∂Qc ∂n
]µ
⎡ ∂Q ⎤ ⎡ ∂Q ⎤ Var (n) + ⎢ c ⎥ Var ( D) + ⎢ c ⎥ Var ( S ) ⎣ ∂D ⎦ µ ⎣ ∂S ⎦ µ 2
2
(− 0.463D + (0.463n
2.67
−1
S 0.5 D −2
)
2
µ
2
(
)
2
Var (n) + 0.463n −1 × 2.67 D 1.67 S 0.5 µ Var ( D)
)
2
D 2.67 × 0.5S −0.5 µ Var ( S )
=22.665(ft3/sec)2 (b) The percentage of contribution of each random variable to the total uncertainty:
∂Qc ∂n
[
]µ 2 Var (n) / Var (Qc ) = 74.2%
⎡ ∂Qc ⎤ ⎢ ∂D ⎥ Var ( D) / Var (Qc ) = 21.2% ⎣ ⎦µ 2
⎡ ∂Qc ⎤ ⎢ ∂S ⎥ Var ( S ) / Var (Qc ) = 4.6% ⎣ ⎦µ 2
(c) QC follows the log-normal distribution with:
µ Q = 41.0 C
σ
2 QC
= 22.665
σ 22.665 = = 0.116 µ 41 λ = ln µ − 12 ζ 2 = 3.707
⇒ζ ≈
∴ P(QC > 30) = 1 − Φ (
ln 30 − 3.707 ) = 0.9958 0.116
4.37 50
(a)
T= ∑ Ri i =1
µ T = 50 × 1 = 50
σ T = 50 × 12 = 7.07 ∴ P(T > 60) = 1 − Φ ( (b) i.
60 − 50 ) = 1 − Φ (1.414) = 1 − 0.9213 = 0.078 7.07
V = N R ln( R + 1) =1*1*ln(1+1) =ln2 =0.693 ⎛ ∂V ⎞ ⎛ ∂V ⎞ Var(V)= ⎜ ⎟ Var ( R) ⎟ Var ( N ) + ⎜ ⎝ ∂R ⎠ ⎝ ∂N ⎠ 2
[
ii.
2
]
2
R ⎡ ⎤ = R ln( R + 1) (0.5) + ⎢ N ( + ln( R + 1))⎥ × 1 ⎣ R +1 ⎦ 2 2 2 =(ln2) (0.5) +(1/2+ln2) ×1 =0.48*0.25+1.424*1 =0.12+1.424 =1.544 1.544 = 1.793 ∴δ V = 0.693 The approximation is not expected to be good because: 2
2
a)
R and N have large variance; b) The function is highly nonlinear Check the second order term for the mean:
1 ∂ 2V 1 ⎡ ( R + 1) − R 1 ⎤ ⋅ Var ( R) = ⎢ N ( )⎥Var ( R ) = 0.375 + 2 2 2 ∂R 2⎣ R +1 ⎦ ( R + 1) which is not that small compared to the first order term. ∴the approximations are not good.
w=
K M σK = 200
µK = 400,
M = 100,
The first order mean and variance of w are
1 1 µk = 400 = 2 10 10 1 1 1 1 σ w2 ≈ σ K2 ( µ K − 2 ) 2 = 200 2 ( ) 2 = 0.25 10 2 2 400
µw ≈
Hence,
σw = 0.5
M is also random with µM = 100 and σM = 20 The first order mean and variance of w are
µw ≈
µK 400 = =2 100 µM 1 2
1 2
σ w2 ≈ 0.25 + σ M2 ( µ K µ M −3 / 2 ) 2 = 0.25 + 20 2 ((20)(100) −3 / 2 ( )) 2 = 0.25 + 0.04 = 0.29 Hence, σw = 0.538 The second order approximate mean of w is
1 2 ∂2w 1 2 ∂2w µw ≈ 2 + σ K ( 2 )µ + σ M ( )µ 2 2 ∂K ∂M 2 1 1 1 1 1 3 −3 / 2 µ M −1 / 2 + σ M2 (− )(− ) µ K 1 / 2 µ M − 5 / 2 = 2 + σ K2 (− )( ) µ K 2 2 2 2 2 2 1 3 = 2 − (200) 2 (400) − 3 / 2 (100) −1 / 2 + (20) 2 (400)1 / 2 (100) − 5 / 2 8 8 = 2 – 0.0625 + 0.03 = 1.968
6.1
∑x
∧ 765 = 85 ton = µ x n 9 ( xi − 85) 2 366 ∑ 2 = = 45.75 s = n −1 8
(a) x =
i
=
∧
σ x = 45.75 = 6.7643 (b) one-sided hypothesis test Null hypothesis Ho: Alternative hypothesis HA:
µ = 80 tons µ < 80 tons
Without assuming known standard deviation, the estimated value of the test statistic to be
t=
x − 80 s/ n
=
85 − 80 6.764 / 9
= 2.2176
with f = 9-1 = 8 d.o.f. we obtain the critical value of t from Table A.2. 2 at the 5% significance level to be tα = -1.8595. Therefore the value of the test statistic is 2.2176 > 1.8595 which is outside the region of rejection; hence the null hypothesis is accepted, and the capacity of the pile remain acceptable. (c) < µ x >0.98
= ( x - k 0.01
σx
, x + k 0.01
σx
)
n n 6.764 6.764 , 85 + 2.33 ) = ( 85 – 2.33 3 3 = (79.747, 90.253)
(d) < µ x >0.98
sx s , x + t 0.01, 8 x ) n n 6.764 6.764 , 85 + 2.8965 ) = ( 85 – 2.8965 3 3 = ( x - t 0.01, 8
= (78.47, 91.53)
6.2 (a) Since x = 65, n = 50, σ = 6, a 2-sided 99% interval for the true mean is given by the limits 6 65 ± k0.005 50 ⇒ <µ>0.99 = (65 – 2.58×
6
, 65 – 2.58×
50
6
)
50
= (62.81, 67.19) (in mph) 6 2 (b) Requiring 2.58 ≤ 1 ⇒ n ≥ (2.58×6) = 239.6304 n ⇒ n = 240 ⇒ (240 – 50) = 190 additional vehicles must be observed. Before we proceed to (c) and (d), let X J and X M denote John and Mary’s sample means, respectively. Both are approximately normal with mean µ (unknown true mean speed) and 6 standard deviation , hence their difference n D = XJ – XM has an (approximate) normal distribution with mean = µ – µ = 0, and 2
2
⎛ 6 ⎞ ⎛ 6 ⎞ 2 ⎟⎟ = 6 ⎟⎟ + ⎜⎜ , standard deviation = ⎜⎜ n ⎝ n⎠ ⎝ n⎠ i.e. D ~ N(0, 6
2 ) n
Hence (c) When n = 10, P( X J – X M > 2) = P(D > 2) D − µD 2−0 = P( > ) σD 6/ 5 = 1 – Φ(0.745) ≅ 0.228 (d) When n = 100, P(D > 2) = 1–Φ (
2−0 ) = 1 – Φ( 2.357) ≅ 0.0092 6 / 50
6.3 (a) The formula to use is <µ>1 - α = [ x − tα / 2,n −1
s n
, x − tα / 2,n −1
s
], where
n
α = 0.1, n = 10, x = 10000 cfs, s = 3000 cfs, and tα / 2,n −1 = t 0.05,9 = 1.833 Hence, the 2-sided 90% confidence interval for the mean annual maximum flow is [(10000 - 1739.044) cfs, (10000 + 1739.044) cfs] ≅ [8261 cfs, 11739 cfs] (b) Since the confidence level (1 - α) is fixed at 90%, while s is assumed to stay at approximately s 3000 cfs, the “half-width” of the confidence interval, tα / 2, n −1 can be considered as a n function of the sample size n only. To make it equal to 1000 cfs, one must have t 0.05,n −1 ⇒
3000
= 1000
n
t 0.05, n −1
= 1/3, n which can be solved by trial and error. We start with n = 20 (say), and increase the sample size until the half width is narrowed to the desired 1000cfs, i.e. until t 0.05, n −1 ≤ 0.33333…. n With reference to the following table,
Sample size, n
t0.05,n-1
20 21 22 23 24 25 26 27
1.7291 1.7247 1.7207 1.7171 1.7138 1.7108 1.7081 1.7056
t 0.05, n −1 n 0.387 0.376 0.367 0.358 0.350 0.342 0.335 0.328
We see that a sample size of 27 will do, hence an additional (27 – 10) = 17 years of observation will be required.
6.4 (a) Pile Test No. N = A/P 5
(b) N =
∑N i =1
1
2
3
4
5
1.507
0.907
1.136
1.070
1.149
5
i
∑ (N
i
− N )2
≅ 1.154, = i =1 ≅ 0.219850903 ≅ 0.048 5 5 −1 (c) Substituting n = 5, N = 1.15392644, SN = 0.219850903 into the formula S <µN>0.95 = N ∓ t0.025, n – 1 N , n where t0.025, n – 1 = t0.025,4 ≅ 2.776, we have (1.15392644 ∓ 0.27293719) = (0.881, 1.427) as a 95% confidence interval for the true mean of N.
S N2
2
(d) With σ = 0.045 known, and assuming N is normal, to estimate µN to ±0.02 with 90% confidence would require k0.05
0.045
≤ 0.02
n
⇒ n ≥ (1.645
0.045 2 ) = 304.37 0.02
⇒ n = 305, hence an additional (305 – 5) = 300 piles should be tested.
(e) Since A = 15N, P(pile failure) = P(A < 12) = P(15N < 12) = P(N < 12/15) =Φ(
12 / 15 − 1.154 ) 0.22
= Φ(–1.61) ≅ 0.0537
6.5 5
(a) Let x be the concrete strength. x =
∑x i =1
5
5
i
= 3672, Sx =
∑ (x i =1
i
− x) 2
5 −1
=
589778 , hence 4
589778 4 <µN>0.90 = 3672 ∓ t0.05, 5 – 1 , 5 where t0.05,4 ≅ 2.131846486, we have (3672 ∓ 366.09) = (3305.91, 4038.09) as a 90% confidence interval for the true mean of N. Note: our convention is that the subscript p in tp, dof always indicates the tail area to the right of tp, dof
(b) If the “half-width” of the confidence interval, tα/2,
4
589778 4 is only 300 (psi) wide, it 5
implies 589778 4 ) = 1.746996, tα/2, 4 = 300 / ( 5 We need to determine the corresponding α. On the other hand, from t-distribution table we have (at 4 d.o.f.) t 0.1, 4 = 1.533, t 0.05, 4 = 2.132, (and tα/2, 4 increases as α/2 gets smaller in general) Hence we may use linear interpolation to get an approximate answer: over such a small “x” range (from t = 1.533 to t = 2.132), we treat “y” (i.e. α/2) as decreasing linearly, with slope m=
0.05 − 0.1 = – 0.083472454 2.132 − 1.533
Hence, as t goes from 1.533 to 1.746996, α/2 should decrease from 0.1 to α/2 = 0.1 + m(1.746996 – 1.533) = 0.1 – 0.083472454×0.213996232 = 0.082137209 ⇒ α = 2×0.082137209 = 0.164274419 Hence the confidence level is 1 – α = 1 – 0.164274419 ≅ 83.6%
6.6 (a) N = 30, x = 12.5 tons Assume σ = 3 tons Assume σ = 3 tons < µ x >0.99
= ( x - k 0.005
σ n
= ( 12.5 – 2.58 ×
, x + k 0.005
3
σ
)
n
, 85 +12.5 + 2.58 ×
30
3
)
30
= (12.5 – 2.826, 12.5 – 2.826) = (9.674, 15.326) tons (b) Let n’ be the total number of observations required for estimating to ± 1 tone with 99% confidence. Hence
2.58 × or
3 ≡1 n'
n’ = (2.58x3)2 = 59.9
Therefore, (60 - 30) or 30 additional observations of truck weights would be required.
6.7 fH(h) =
2h
α
2
−h2
eα
2
From Eq. 6.4, the likelihood function of n observations on a Rayleigh distributinon is n
L=
2hi
∏α i =1
2
− hi 2
eα
2
Taking logarithm of both sides ln L = n ln 2 – 2n ln α + Σ ln hi –
∂ ln L 2n 2Σhi =− + =0 α α3 ∂α
1
α2
Σ hi2
2
Æ
From the given data, n = 10, Σ hi2 = (1.52 + 2.82 + …… 2.32) = 80.42 ∧
∴ α = ± 2.836
n
where
∑ ∑ =
i =1
∧2
α =
∧ Σhi Σhi or α = ± n n 2
2
6.8 (a) x = 2.03 mg/l, s = 0.485 mg/l One-side hypothesis test that the minimum concentration DO is 2.0 mg/l . µ = 2.0 mg/l µ > 2.0 mg/l
Null hypothesis Ho: Alternative hypothesis HA:
Without assuming known standard deviation, the estimated value of the test statistic to be
x−2
t=
0.485 / 10
=
2.03 − 2
= 0.0196
0.485 / 10
with f = 10-1 = 9 d.o.f. we obtain the critical value of t from Table A.2. 2 at the 5% significance level to be tα =1.833. Therefore the value of the test statistic is 0.0196 <1.833, which is outside the region of rejection; hence the stream quality satisfy the EPA standard. (b) < µ >0.95
s s , DO + t 0.025, 9 ) 10 10 0.485 0.485 , 2.03 + 2.2622 ) = ( 2.03 – 2.2622 10 10 = ( DO - t 0.025, 9
= (1.683, 2.377) mg/l (c) < µ >0.95
= DO - t 0.05, 9
s 0.485 = 2.03 – 1.8331 =1.749mg/l 10 10
6.9
124.3 + 124.2 + 124.4 = 124.3 ft. 3 1 2 SL = [(124.3 − 124.3) 2 + (124.2 − 124.3) 2 + (124.4 − 124.3) 2 ] = 0.01 3 −1 s 0.01 = 0.0577 ft. Standard error of the mean = L = n 3 40o 24.6'+...... + 40o 25.2' = 40o 25' (b) β = 5 1 2 Sβ = [(40o 24.6'−40o 25.0' ) 2 + ...... + (40o 25.2'−40o 25.0' ) 2 ] = 0.135' 5 −1 sβ 0.135 = = 0.164' = 4.77 × 10− 5 radian Standard error of the mean = n 5
(a) L =
(c) h = 3 ft ;σ h = 0.01 ft H = h + L tan β
H = h + L tan β = 3 + 124.3 × tan 40o 25' = 108.85 ft. (d) σ H 2 = Var ( h ) + Var ( L )(tan β )2 + Var( β ) x ( L •sec2 β )2 = (0.01)2 + (0.0577)2(tan 40o25’)2 + (4.77 x 10-5)2 x (124.3 sec2 40o25’)2 = 0.0001+0.0024+0.0001 = 0.0026
σ H = 0.051 ft. (e) P (− k α / 2 <
H −H
σH
≤ k α / 2 ) = 0.98
108.85 − H ≤ 2.33) = 0.98 0.051 or P (108.73 ≤ H < 108.97) = 0.98 or P (−2.33 <
So
0.98 = (108.73, 108.97) ft
6.10
2.5 + 2.4 + 2.6 + 2.6 + 2.4 = 2.5cm 5 1.6 + 1.5 + 1.6 + 1.4 + 1.4 r2 = = 1.5cm 5 1 2 S r1 = [(2.5 − 2.5) 2 + (2.4 − 2.5) 2 × 2 + (2.6 − 2.5) 2 × 2] = 0.01 5 −1 S r1 = 0.1cm ; S r 1 = 0.01 / 5 = 0.0447
(a) r1 =
Similarly, S r 2 = 0.01 / 5 = 0.0447 (b) A = π ( r 12 – r 22) = π (2.52 – 1.52) = 12.57cm2 (c) σ A 2 = S r 1 2 (2π r 1)2 + S r 2 2 (-2π r2 )2
σA
= (0.0447)2 (2π x 2.5)2 + (0.0447)2 (-2π x 1.5)2 = 0.671 = 0.819cm2
(d) tα / 2, n −1
sr1 n
= 0.07
S r1 = 0.1 So, n = (
tα / 2, n −1 2 ) ------ (1) 0 .7
Assume n=10, tα/2,9 = 3.2498 so from Equation (1), n = 21.55. For n = 16, tα/2,15 = 2.9467 and n = 17.7 For n = 17, tα/2,16 = 2.9208 and n = 17.4 For n = 18, tα/2,17 = 2.8982 and n = 17.14 From trial and error, n is found to be 17. So additional measurements are to be made = 17-5 = 12.
6.11 (a) From the observation data we can get a = 119.9375 , b = 449.5 , β = 59.9583 , (b) sa2 = 0.131 , sb2 = 0.15 , sβ2 = 0.1394% (c) The mean area of the triangle is
A=
1 ab sin β = 0.5 × 119.9375 × 449.5 × sin 59.9583 = 23334.73m 2 2
The variance of the area is:
1 1 1 2 2 2 = 0.5 × 449.5 × sin 59.9583 × 0.131 + 0.5 × 119.9375 × sin 59.9583 × 0.15
σ A2 = b sin βσ a2 + a sin βσ b2 + ab cos βσ β2
+0.5 × 119.9375 × 449.5 × cos59.9583 × 0.1394% = 52.086m 4
σ A = 7.217m 2 (d)
A
0.9
= ( A + k0.05 iσ A , A + k0.95 iσ A ) = ( 23334.73 − 1.64 × 7.217, 23334.73 + 1.64 × 7.217 ) = ( 23322.89,23346.57 )
6.12 (a) From the observation data a = 80.06 , sa2 = 0.182 ; b = 120.52 , sb2 = 0.277 Thus A =
1 ab = 0.5 × 80.06 × 120.52 = 4824.42m 2 2
(b)
1 1 2 2 2 σ A = 4.696m
σ A2 = bσ a2 + aσ b2 = 0.5 × 120.52 × 0.182 + 0.5 × 80.06 × 0.277 = 22.056m 4 (c)
A
0.95
= ( A + k0.025 iσ A , A + k0.975 iσ A ) = ( 4824.42 − 1.96 × 4.696, 4824.42 + 1.96 × 4.696 ) = ( 4815.21,4833.62 )
(d) It can be shown that
E (C ) = E A [ E (C | A)] = 10000 + 15 × E ( A) = 10000 + 15 × 4824.42 = 82366.3$
and
Var (C ) = E A [Var (C | A) ] + VarA [ E (C | A)] = 200002 + VarA [10000 + 15 A] = 200002 + 152 × σ A2 = 200002 + 152 × 22.056 = 400004962.6$2 σ C = 20000.12$ P(C > 90000) = 1 − P(C < 90000) = 1 − Φ (
90000 − 82366.3 ) = 0.35 20000.12
6.13 Let x denotes the mileage (a) x = 37.4 s x = 4.06 (b) One-sided hypothesis test that the stated mileage is smaller than 35 mpg: Null hypothesis Ho: µ = 35 mpg Alternative hypothesis HA: µ < 35 mpg Without assuming known standard deviation, the estimated value of the test statistic to be
t=
37.4 − 35 x − 35 = = 1.87 4.06 / 10 4.06 / 10
with f = 10-1 = 9 d.o.f. we obtain the critical value of t from Table A.2. 2 at the 2% significance level to be tα =-2.448. Therefore the value of the test statistic is 1.87>-2.448, which is outside the region of rejection; hence the stated mileage is satisfied. (c)
x
0.95
s s ⎞ ⎛ = ⎜ x − t0.025,9 i x , x + t0.025,9 i x ⎟ 10 10 ⎠ ⎝ 4.06 4.06 ⎞ ⎛ = ⎜ 37.4 − 2.262 × ,37.4 + 2.262 × ⎟ 10 10 ⎠ ⎝ = ( 34.50,40.30 )
7.1 Median = 76.9 c.o.v. = ln (81.5 / 76.9) = 0.06
7.2 (a) The CDF of Gumbel distribution is:
FX ( x ) = exp ( − e −α ( x − µ ) ) , −∞ < x < ∞
The standard variate is S = −α ( X − µ )
(
With CDF of: FS ( s ) = exp − e − s
)
If x is observed with frequency of p , then its corresponding standard variate is
s = − ln ( − ln( p ) ) Thus, the observed data can be arranged in the right table and plotted in the following figure.
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20
x 69.3 70.6 72.3 72.9 73.5 74.8 75.8 75.9 76 76.1 76.4 77.1 77.4 78.2 78.2 78.4 79.3 80.8 81.8 85.2
P 0.048 0.095 0.143 0.190 0.238 0.286 0.333 0.381 0.429 0.476 0.524 0.571 0.619 0.667 0.714 0.762 0.810 0.857 0.905 0.952
s -1.113 -0.855 -0.666 -0.506 -0.361 -0.225 -0.094 0.036 0.166 0.298 0.436 0.581 0.735 0.903 1.089 1.302 1.554 1.870 2.302 3.020
90
Standard variate X
80 70 60 50
X = 3.3942S + 74.723
40 30 20 10 0 -2.000 -1.000 0.000
1.000
2.000
3.000
4.000
Observed maximum wind velocity V
(b) The linear regression of the data gives the trend line equation of
X = 3.3942 S + 74.723
which gives
X − 74.723 = 0.295( X − 74.723) 3.3942 Thus, α = 0.295 , µ = 72.723 S=
(c) Perform Chi-Square test for Gumbel distribution Interval
65-70 70-75 75-80 80-85
Σ
Observed frequency ni
Theoretical frequency ei
( ni − ei ) 2
( ni − ei ) 2 ei
1 5 11 3 20
0.356 7.602 8.240 2.860 20
0.415 6.770 7.617 0.020
1.164 0.891 0.924 0.007 2.986
The degree of freedom for the Gumbel distribution is f=4-1-2=1. For a significance level α = 2% , c.98,1 = 5.41 . Comparing with the Σ ( ni − ei ) 2 ei calculated, the Gumbel distribution is a valid model for the maximum wind velocity at the significance level of α = 2% .
7.3 (a) m
εµ
m N +1
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
16.0 16.1 16.6 16.8 17.0 17.3 17.8 17.9 18.1 18.4 18.6 18.8 19.1 19.4 20.1
0.0625 0.1250 0.1875 0.2500 0.3125 0.3750 0.4375 0.5000 0.5625 0.6250 0.6875 0.7500 0.8125 0.8750 0.9375
From Figs. 7.3a and 7.3b, it can be observed that both models (normal and lognormal) fit the data pretty well. (b)
U = 17.8 sU2 = 1.312
σ2 1.312 2 ζ = ln(1 + 2 ) = ln(1 + ) = 0.074 µ 17.82 1 2
λ = ln µ − ζ 2 = ln(17.8) − 0.5 × 0.0742 = 2.876 Perform Chi-square test for normal distribution: Observed Theoretical frequency ni frequency ei 15.5-16.5 16.5-17.5 17.5-18.5 18.5-19.5 19.5-20.5
∑
3.000 3.000 4.000 4.000 1.000 15
1.816 3.730 4.404 2.989 1.166 15
Perform Chi-square test for lognormal distribution: Observed Theoretical Interval frequency ni frequency ei 15.5-16.5 16.5-17.5
3.000 3.000
1.939 3.943
( ni − ei ) 2 1.401 0.533 0.163 1.021 0.028
( ni − ei ) 2 ei 0.772 0.143 0.037 0.342 0.024 1.317
( ni − ei ) 2
( ni − ei ) 2 ei
1.126 0.890
0.581 0.226
17.5-18.5 18.5-19.5 19.5-20.5
∑
4.000 4.000 1.000 15
4.316 2.778 1.133 15
0.100 1.493 0.018
0.023 0.537 0.016 1.382
The degree of freedom for the Normal and Lognormal distribution are both f=5-1-2=2.
( ni − ei ) 2 ∑ e calculated, i the 2 distributions are both valid for the rainfall intensity at the significance level of α = 2% . ( ni − ei ) 2 As the ∑ in Normal distribution is less than that of the Lognormal distribution, the ei
For a significance level α = 2% , c.98,2 = 7.99 . Comparing with the
Normal distribution is superior to the Lognormal distribution in this problem. (c) Perform the K-S test as follows: F D F D (Normal) (Normal) (Lognormal) (Lognormal) k x S 1 15.8 0.067 0.064 0.003 0.059 0.008 2 16.0 0.133 0.085 0.048 0.081 0.052 3 16.1 0.200 0.098 0.102 0.095 0.105 4 16.6 0.267 0.180 0.086 0.184 0.083 5 17.0 0.333 0.271 0.062 0.282 0.052 6 17.3 0.400 0.352 0.048 0.366 0.034 7 17.8 0.467 0.500 0.033 0.517 0.051 8 17.9 0.533 0.530 0.003 0.547 0.014 9 18.1 0.600 0.590 0.010 0.606 0.006 10 18.4 0.667 0.676 0.010 0.688 0.022 11 18.6 0.733 0.729 0.004 0.738 0.005 12 18.8 0.800 0.777 0.023 0.783 0.017 13 19.1 0.867 0.839 0.028 0.840 0.026 14 19.4 0.933 0.889 0.045 0.886 0.047 15 20.1 1.000 0.960 0.040 0.954 0.046 max 0.102 0.105 The K-S test shows that the Normal distribution is more suitable in this case as it gives smaller maximum discrepancy.
Fig. 7.3a
Fig. 7.3b
7.4 (a)
(b)
The minimum time-to-failure = 0 hr. Mean time to failure =
1
λ
= 670 hrs.
(c) Time-to-failure (hours) 0 – 400 400 – 800 800 – 1200 > 1200 Σ
Observed Frequency ni 20 7 7 6 40
Theoritical Frequency ei 17.98 9.90 5.45 6.67 40
(ni-ei)2
(ni-ei)2 / ei
4.08 8.41 2.40 0.45
0.227 0.849 0.441 0.067 1.584
x2 distribution with f = 4 - 2 = 2 degrees of freedom and α = 5% significance level, c0.95, 2 = 5.99
(ni − ei ) 2 = 1.584 < 5.99 ∑ ei i =1 4
In this case
Hence, the exponential distribution is a valid distribution model.
(d) Perform K-S test as follows: k 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 max Where t
x 0.13 0.78 2.34 3.55 8.82 14.29 29.75 39.1 54.85 62.09 84.09 85.28 121.58 124.09 151.44 163.95 216.4 298.58 380 393.37
S 0.025 0.050 0.075 0.100 0.125 0.150 0.175 0.200 0.225 0.250 0.275 0.300 0.325 0.350 0.375 0.400 0.425 0.450 0.475 0.500
F(Normal) D(Normal) k x S F(Normal) D(Normal) 0.000 0.025 21 412.03 0.525 0.462 0.063 0.001 0.049 22 470.97 0.550 0.507 0.043 0.004 0.071 23 633.98 0.575 0.615 0.040 0.005 0.095 24 646.01 0.600 0.621 0.021 0.013 0.112 25 656.04 0.625 0.627 0.002 0.021 0.129 26 658.38 0.650 0.628 0.022 0.044 0.131 27 672.87 0.675 0.636 0.039 0.057 0.143 28 678.13 0.700 0.639 0.061 0.079 0.146 29 735.89 0.725 0.669 0.056 0.089 0.161 30 813 0.750 0.705 0.045 0.119 0.156 31 855.95 0.775 0.724 0.051 0.120 0.180 32 861.93 0.800 0.726 0.074 0.167 0.158 33 862.93 0.825 0.727 0.098 0.170 0.180 34 895.8 0.850 0.740 0.110 0.204 0.171 35 952.65 0.875 0.761 0.114 0.218 0.182 36 1057.57 0.900 0.796 0.104 0.278 0.147 37 1296.93 0.925 0.858 0.067 0.362 0.088 38 1407.52 0.950 0.880 0.070 0.435 0.040 39 1885.22 0.975 0.941 0.034 0.446 0.054 40 2959.47 1.000 0.988 0.012 0.182 2 = 541.19 st = 359937.2 The sample size is 40. At the 5% significance level,
0.05 D40 = 0.21 . As the maximum discrepancy is smaller than the critical value, the exponential
distribution is verified.
7.5 (a) From the observation data we can get v = 1.08 , sv2 = 1.243 (b) Total no. Nos. of Observed of veh. veh/min Freq. Observed ni 0 6 0 1 8 8 6 2x3 + 3x2 ≥2 + 4x1 = 16 Σ20 Σ24
ν=
Theoretical Probability (v = 1.2) 0.301 0.361 0.338
Theoretical Frequency ei 6.02 7.22 6.76
Σ1.0
Σ20.0
24 = 1.2 20
(ni-ei)2
(ni-ei)2 / ei
0.0004 0.6084 0.5776
0.0001 0.0843 0.0854
x2 distribution with f = 3-2 = 1 degree of freedom and α = 1% significance level, c0.99,1 = 6.635
(n i − ei ) 2 In this case ∑ = 0.1698 < 6.635 ei i =1 3
Hence, the Poisson distribution is a valid distribution model.
Σ0.1698
7.6 (a)
f X ( x) =
S=
x
α2
e
1 x − ( )2 2 α
≥0
X
α dX =α dS
X = Sα ,
So f S ( s ) = f X ( sα ) = s ⋅e =0
FS ( s ) = 1 − e
1 − s2 2
1
− s2 dX = s ⋅e 2 dS
;
s≥0
;
s<0
1 − s2 2
s 0.46 0.67 0.84 1.01 1.18 1.26 1.35 1.45 1.55 1.67 1.79 1.95 2.15
FS(s) 0.10 0.20 0.30 0.40 0.50 0.55 0.60 0.65 0.70 0.75 0.80 0.85 0.90
s 2.25 2.31 2.37 2.45 2.49 2.54 2.59 2.65 2.72 2.80 2.90 3.03 3.26
FS(s) 0.92 0.93 0.94 0.95 0.955 0.96 0.965 0.97 0.975 0.98 0.985 0.99 0.995
The corresponding probability paper is shown in Fig. 7.6. The slope of a straight line on this paper is,
dX =α ds which is the model value of X.
Fig. 7.6
(c)
On the basis of the observed data, Rayleigh distribution does not appear to be a suitable model for live-load stress range in highway bridges. However, if the observed data indeed fell on a straight line, the slope of the fitted line will give the most probable stress range.
7.7 .(a) The middle point of each range is used to represent the range. No. of Ki observation( ni ) ni Ki 0.025 0.075 0.125 0.175 0.225 0.275 0.325
1 11 20 23 15 11 2 83
Σ
0.025 0.825 2.500 4.025 3.375 3.025 0.650 14.425
( Ki − K ) 2
ni ( Ki − K ) 2
0.022 0.010 0.002 0.000 0.003 0.010 0.023
0.022 0.107 0.048 0.000 0.039 0.113 0.046 0.375
sK2 = 0.0046
K = 0.174 (b) K (per day) < 0.100 0.100 – 0.150 0.150 – 0.200 0.200 – 0.250 > 0.250 Σ
Observed frequency ni 12 20 23 15 13 83
Theoretical frequency ei 11.09 19.06 24.56 18.25 10.04 83
(ni-ei)2
(ni-ei)2 / ei
0.8281 0.8836 2.4336 10.5625 8.7616
0.075 0.046 0.099 0.579 0.873 1.672
x2 distribution with f = 5-3 = 2 degree of freedom and α = 5% significance level, c0.95,2 = 5.991
(n i − ei ) 2 In this case ∑ = 1.672 < 5.991 ei i =1 5
So the normal distribution is acceptable.
7.8 (a)
X −a dX ; X = Sr + a; =r r dS dX 2 So f s ( s ) = f x ( sr + a ) = 2 ( sr ) ⋅ r dS r = 2s ; 0 ≤ s ≤ 1 S=
=0 (b)
; elsewhere
FS(s) = S2 ; 0 ≤ s ≤ 1 = 1 ; S > 1.0 =0 ; S<0 s 0.224 0.316 0.447 0.548 0.632 0.707 0.742 0.775 0.806 0.837 0.866 0.894 0.922 0.949
FS(s) 0.05 0.10 0.20 0.30 0.40 0.50 0.55 0.60 0.65 0.70 0.75 0.80 0.85 0.90
s 0.959 0.964 0.970 0.975 0.977 0.980 0.982 0.985 0.987 0.990 0.992 0.995
FS(s) 0.92 0.93 0.94 0.95 0.955 0.96 0.965 0.97 0.975 0.98 0.985 0.99
The corresponding probability paper is shown in Fig 7.8.
X −a → x=a r X −a → x =r+a When s = 1 = r ∴ values of X at FS(0) and FS(1.0) correspond to a and r+a which are the minimum and When s = 0 =
maximum values of X. (c) m 1 2 3 4 5 6 7 8
X 18 28 32 34 36 45 48 50
m N +1 0.045 0.091 0.136 0.182 0.227 0.273 0.318 0.364
9 10 11 12 13 14 15 16 17 18 19 20 21
53 53 55 56 58 62 64 66 69 71 71 72 75
0.409 0.455 0.500 0.545 0.591 0.636 0.682 0.727 0.773 0.818 0.864 0.909 0.955
Fig. 7.8
7.9
m 1 2 3 4 5 6 7 8 9 10 11 12 13
shear strength (S) (Ksf) 0.35 0.40 0.41 0.42 0.43 0.48 0.49 0.58 0.68 0.70 0.75 0.87 0.96
m N +1 0.0714 0.1429 0.2143 0.2857 0.3571 0.4286 0.5000 0.5714 0.6429 0.7143 0.7857 0.8571 0.9286
7.10 (a)
Number of Observatio
250 200 150 100 50 0 1
2
3
4
5
6
7
8
9
10 11 12
Acceptance Gap G, (sec)
(b)
The middle point of each range is used to calculated the sample mean and sample variance as follows:
Gi 1 2 3 4 5 6 7 8 9 10 11 12
Σ
No. of observation( ni )
niGi
(Gi − G ) 2
ni (Gi − G ) 2
0 6 34 132 179 218 183 146 69 30 3 0 1000
0 12 102 528 895 1308 1281 1168 621 300 33 0 6248
27.542 18.046 10.550 5.054 1.558 0.062 0.566 3.070 7.574 14.078 22.582 33.086
0.000 108.273 358.683 667.063 278.793 13.408 103.487 448.148 522.572 422.325 67.745 0.000 2990.496
sG2 = 2.993
G = 6.248 Perform the Chi-square test for normal distribution Observed Theoretical Interval frequency ni frequency ei 0.5-1.5 1.5-2.5 2.5-3.5
0 6 34
2.586 12.113 40.966
( ni − ei ) 2
( ni − ei ) 2 ei
6.688 37.368 48.520
2.586 3.085 1.184
3.5-4.5 4.5-5.5 5.5-6.5 6.5-7.5 7.5-8.5 8.5-9.5 9.5-10.5 10.5-11.5 11.5-12.5
132 179 218 183 146 69 30 3 0
∑
100.063 176.577 225.150 207.452 138.121 66.443 23.088 5.794 1.050
1019.948 5.870 51.116 597.887 62.074 6.537 47.774 7.804 1.102
1000
25.204
Perform the Chi-square test for lognormal distribution Observed Theoretical Interval frequency ni frequency ei 0.5-1.5 1.5-2.5 2.5-3.5 3.5-4.5 4.5-5.5 5.5-6.5 6.5-7.5 7.5-8.5 8.5-9.5 9.5-10.5 10.5-11.5 11.5-12.5
0 6 34 132 179 218 183 146 69 30 3 0
∑
10.193 0.033 0.227 2.882 0.449 0.098 2.069 1.347 1.050
0.000 0.610 22.354 119.016 227.509 241.300 179.619 107.248 55.622 26.330 11.745 5.044
( ni − ei ) 2
( ni − ei ) 2 ei
0.000 29.047 135.621 168.591 2353.093 542.883 11.434 1501.746 178.971 13.472 76.484 25.441
0.000 47.582 6.067 1.417 10.343 2.250 0.064 14.003 3.218 0.512 6.512 5.044
1000
97.009
Where the lognormal distribution parameter is calculated as follows:
1 2
λ = ln µ − ζ 2 = ln(6.248) − 0.5 × 2.993 = 1.795 ζ = ln(1 +
σ2 2.993 ) = ln(1 + ) = 0.272 2 µ 6.2482
As 25.204<97.009, the normal distribution is more suitable for this problem. (c) Acceptance gap size (secs) Gi 1 2 3 4 5
No. of Observations ni
Gi ⋅ ni ∑ ni
(Gi-µG)2
(Gi-µG)2 ni / Σni (x10-3)
0 6 34 132 179
0 0.012 0.104 0.528 0.895
27.56 18.0625 10.5625 5.0625 1.5625
0 108.375 359.125 668.250 279.688
6 7 8 9 10 11 12 Σ µG = 6.25 sec Var(G) = 2.99 σ G = 1.729
218 183 146 69 30 3 0 1000
1.308 1.281 1.168 0.621 0.300 0.033 0 6.25
0.0625 0.5625 3.0625 7.5625 14.0625 22.5625 33.0625
13.625 102.938 447.125 521.813 421.875 67.688 0 2990.502x10-3
7.11 Rearrange the given data in increasing order, we obtain the following table: Observed Rainfall m m/N+1 S Intensity X, in 1 39.91 0.03 -1.83 2 40.78 0.07 -1.50 3 41.31 0.10 -1.28 4 42.96 0.13 -1.11 5 43.11 0.17 -0.97 6 43.30 0.20 -0.84 7 43.93 0.23 -0.73 8 44.67 0.27 -0.62 9 45.05 0.30 -0.52 10 45.93 0.33 -0.43 11 46.77 0.37 -0.34 12 47.38 0.40 -0.25 13 48.21 0.43 -0.17 14 48.26 0.47 -0.08 15 49.57 0.50 0.00 16 50.37 0.53 0.08 17 50.51 0.57 0.17 18 51.28 0.60 0.25 19 53.02 0.63 0.34 20 53.29 0.67 0.43 21 54.49 0.70 0.52 22 54.91 0.73 0.62 23 55.77 0.77 0.73 24 58.71 0.80 0.84 25 58.83 0.83 0.97 26 59.12 0.87 1.11 27 63.52 0.90 1.28 28 67.59 0.93 1.50 29 67.72 0.97 1.83 (a) Plot the data points in a lognormal probability paper
100.0
Observed Rainfall Intensity, in
y = 50.16e0.159x
10.0
1.0 -3.00
-2.00
-1.00
0.00
1.00
2.00
3.00
Standard Normal Variate, S
(b) From the data in probability paper xm = 50.16 ⇒ λ = ln(xm) = 3.92 x0.84 = 58.8 ⇒ ζ = ln(x0.84/xm) = 0.159 Perform Chi-square test for log-normal distribution Interval (in)
Observed frequency ni
Theoretical frequency ei
(ni-ei)2
(ni − ei ) 2 ei
<40 40-45 45-50 50-55 55-60 >60
1 7 7 7 4 3
2.118 4.784 7.018 6.629 4.495 3.956
1.250 4.913 0.000 0.137 0.245 0.915
0.590 1.027 0.000 0.021 0.054 0.231
Σ
29
29
1.92
The degree of freedom for the lognormal distribution is f=6-1-2=3. On the basis of the observation data
µˆ = 50.354 σˆ 2 = 61.179 , thus νˆ = 0.823 kˆ = 41.445 Perform Chi-square test for Gamma distribution Observed Theoretical Interval (in) frequency ni frequency ei <40 40-45
1 7
2.454 4.947
( ni − ei ) 2
( ni − ei ) 2 ei
2.116 4.215
0.862 0.852
45-50 50-55 55-60 >60
7 7.173 0.030 7 6.741 0.067 4 4.422 0.178 3 3.263 0.069 29 29 ∑ The degree of freedom for the Gamma distribution is f=6-1-2=3. For a significance level α = 5% , c.95,3 = 7.81 . Comparing with the
0.004 0.010 0.040 0.021 1.790
( ni − ei ) 2 ∑ e calculated, i
both the Gamma distribution and Lognormal appear to be valid model for the rainfall intensity at the significance level of α = 5% . As the
( ni − ei ) 2 ∑ e in Gamma distribution is less than i
that of the lognormal distribution, the Gamma distribution is superior to the lognormal distribution in this problem.
7.12 (a) Plot data on normal probability paper 4.0
Ratio of Settlement
3.0
2.0
y = 0.3322x + 0.994 1.0
0.0 -3.00
-2.00
-1.00
0.00
1.00
2.00
3.00
Standard Normal Variate, S
Linear trend is observed, but there are two data points out of the trend. (b) Based on the trend line, xm = 0.994 ⇒ µ= 0.994 ⇒ σ = slope = 0.3322 (b) Perform Chi-square test for normal distribution, Ratio of settlement
Observed frequency ni
Theoretical frequency ei
(ni-ei)2
<0.75 0.75-1.0 1.0-1.25 1.25-1.5 >1.5 Σ
2 13 8 1 1 25
5.783 6.897 6.808 3.915 1.596425 25
14.312 37.246 1.420 8.499 0.356
(ni − ei ) 2 ei 2.475 5.400 0.209 2.171 0.223 10.25> c.95,2=5.99
The degree of freedom for the normal distribution is f=5-1-2=2. For a significance level α=5%, c.95,2=5.99. Comparing with the
(ni − ei ) 2 ∑ e calculated, the i
normal distribution is not a valid model for the ratio of settlement.
Perform Anderson-Darling test for normal distribution, i
xi
F(xi)
F(xn+1-i)
(2i − 1) {ln FX ( xi ) + ln[1 − FX ( xn+1−i )]} / n
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25
0.12 0.52 0.82 0.84 0.86 0.86 0.87 0.88 0.92 0.94 0.94 0.97 0.99 0.99 1.00 1.01 1.02 1.04 1.04 1.06 1.09 1.14 1.18 1.38 2.37
0.004 0.077 0.300 0.321 0.343 0.343 0.354 0.366 0.412 0.435 0.435 0.471 0.495 0.495 0.507 0.519 0.531 0.555 0.555 0.579 0.614 0.670 0.712 0.877 1.000
1.000 0.877 0.712 0.670 0.614 0.579 0.555 0.555 0.531 0.519 0.507 0.495 0.495 0.471 0.435 0.435 0.412 0.366 0.354 0.343 0.343 0.321 0.300 0.077 0.004
-0.657 -0.560 -0.490 -0.628 -0.727 -0.851 -0.960 -1.089 -1.118 -1.188 -1.293 -1.321 -1.386 -1.447 -1.451 -1.522 -1.536 -1.462 -1.519 -1.509 -1.490 -1.356 -1.253 -0.396 -0.008 Σ=-27.22
Calculate the Anderson-Darling (A-D) statistic n
A2 = −∑ ⎡⎣(2i − 1) {ln FX ( xi ) + ln[1 − FX ( xn +1−i )]} / n ⎤⎦ − n = 27.22-25 = 2.22 i =1
For normal distribution, the adjusted A-D statistic for a sample size n=25 is,
⎛ ⎝
A* = A2 ⎜1.0 +
0.75 2.25 ⎞ + 2 ⎟ =2.294 n n ⎠
The critical value cα is given by,
⎛ b b ⎞ cα = aα ⎜ 1 + 0 + 12 ⎟ n n ⎠ ⎝
= 0.726
for a prescribed significance level α = 5%, aα =0.7514, b0 =-0.795, b1 =-0.89 (Table A.6). As A* is greater than cα , the normal distribution is not acceptable at the significance level α = 5%.
(c) Plot the data on the lognormal probability paper,
Ratio of Settlement
10.0
1.0
y = 0.9143e0.4081x
0.1 -3.00
-2.00
-1.00
0.00
1.00
2.00
3.00
Standard Normal Variate, S
Therefore, xm = 0.9143 ⇒ λ = ln(xm) = -0.0896 ζ = slope = 0.4081 Perform Chi-square test for lognormal distribution, Ratio of settlement
Observed frequency ni
Theoretical frequency ei
(ni-ei)2
<0.75 0.75-1.0 1.0-1.25 1.25-1.5 >1.5 Σ
2 13 8 1 1 25
7.843 6.830 4.784 2.730 2.814 25
34.135 38.073 10.341 2.992 3.290
(ni − ei ) 2 ei 4.353 5.575 2.161 1.096 1.169 Σ= 13.18> c.95,2=5.99
The degree of freedom for the normal distribution is f=5-1-2=2. For a significance level α=5%, c.95,2=5.99. Comparing with the
(ni − ei ) 2 ∑ e calculated, the i
lognormal distribution is not a valid model for the ratio of settlement. Perform Anderson-Darling test for lognormal distribution, i
xi
F(xi)
F(xn+1-i)
(2i − 1) {ln FX ( xi ) + ln[1 − FX ( xn+1−i )]} / n
1 2 3 4
0.12 0.52 0.82 0.84
0.000 0.083 0.395 0.418
0.990 0.843 0.734 0.706
-0.783 -0.521 -0.451 -0.587
5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25
0.86 0.86 0.87 0.88 0.92 0.94 0.94 0.97 0.99 0.99 1.00 1.01 1.02 1.04 1.04 1.06 1.09 1.14 1.18 1.38 2.37
0.440 0.440 0.452 0.463 0.506 0.527 0.527 0.558 0.577 0.577 0.587 0.596 0.606 0.624 0.624 0.641 0.667 0.706 0.734 0.843 0.990
0.667 0.641 0.624 0.624 0.606 0.596 0.587 0.577 0.577 0.558 0.527 0.527 0.506 0.463 0.452 0.440 0.440 0.418 0.395 0.083 0.000
-0.691 -0.812 -0.922 -1.049 -1.096 -1.176 -1.281 -1.330 -1.410 -1.474 -1.487 -1.570 -1.593 -1.530 -1.587 -1.598 -1.617 -1.530 -1.461 -0.484 -0.019 Σ=-28.06
Calculate the Anderson-Darling (A-D) statistic n
A2 = −∑ ⎡⎣(2i − 1) {ln FX ( xi ) + ln[1 − FX ( xn +1−i )]} / n ⎤⎦ − n = 28.06-25 = 3.06 i =1
For lognormal distribution (lognormal distribution should be the same as normal), the adjusted A-D statistic for a sample size n=25 is,
⎛ ⎝
A* = A2 ⎜1.0 +
0.75 2.25 ⎞ + 2 ⎟ =3.16 n n ⎠
The critical value cα is given by,
⎛ b b ⎞ cα = aα ⎜ 1 + 0 + 12 ⎟ n n ⎠ ⎝
= 0.726
for a prescribed significance level α = 5%, aα =0.7514, b0 =-0.795, b1 =-0.89 (Table A.6). As A* is greater than cα , the lognormal distribution is not acceptable at the significance level
α = 5%.
7.13 Plot data on (shifted) exponential probability paper,
100.0
Observed Rainfall Intensity, in
90.0 y = 27.972x - 0.2884
80.0 70.0 60.0 50.0 40.0 30.0 20.0 10.0 0.0 0.00
0.50
1.00
1.50
2.00
2.50
3.00
3.50
4.00
4.50
5.00
Standard Normal Variate, S
⇒ a = -0.2884, λ = 1/27.972=0.03575 Perform Chi-square test for shifted exponential distribution, Mean depth (m)
Observed frequency ni
Theoretical frequency ei
(ni-ei)2
<5 5-15 15-25 25-35 35-45 Σ
3 4 1 4 0 15
2.584 3.732 2.610 1.826 1.277 15
0.173 0.072 2.593 4.728 1.630
(ni − ei ) 2 ei 0.067 0.019 0.993 2.590 1.277 Σ= 4.95 < c.95,2=5.99
The degree of freedom for the shifted exponential distribution is f=5-1-2=2. For a significance level α=5%, c.95,2=5.99. Comparing with the
(ni − ei ) 2 ∑ e calculated, the shifted i
exponential distribution is an appropriate model for the mean depth of Glacier Lakes. Perform Anderson-Darling test for exponential distribution,
i
xi
F(xi)
F(xn+1-i)
(2i − 1) {ln FX ( xi ) + ln[1 − FX ( xn+1−i )]} / n
1 2
2.90 4.70
0.108 0.163
0.950 0.834
-0.348 -0.722
3 4 5 6 7 8 9 10 11 12 13 14 15
5.00 7.10 10.40 12.00 13.60 18.60 27.90 28.60 33.30 34.30 46.90 50.00 83.30
0.172 0.232 0.318 0.356 0.391 0.491 0.635 0.644 0.699 0.710 0.815 0.834 0.950
0.815 0.710 0.699 0.644 0.635 0.491 0.391 0.356 0.318 0.232 0.172 0.163 0.108
-1.149 -1.259 -1.409 -1.516 -1.686 -1.387 -1.077 -1.114 -1.036 -0.931 -0.656 -0.647 -0.320 Σ=-15.26
Calculate the Anderson-Darling (A-D) statistic n
A2 = −∑ ⎡⎣(2i − 1) {ln FX ( xi ) + ln[1 − FX ( xn +1−i )]} / n ⎤⎦ − n = 15.26-15 = 0.26 i =1
For exponential distribution, the adjusted A-D statistic for a sample size n=15 is,
A* = A2 (1.0 +
0.6 ) = 0.3 n
The critical value cα is given in Table A.6b, cα = 1.321 for a prescribed significance level α = 5%. A* < cα , therefore the exponential distribution is acceptable at the significance level α = 5%. Therefore, both Chi-square test and Anderson-Darling test have shown the shifted exponential distribution is an appropriate model for the mean depth of Glacier Lakes.
Acceptance gap size (secs) Gi < 2.5 2.5 – 3.5 3.5 – 3.5 4.5 – 3.5 5.5 – 3.5 6.5 – 3.5 7.5 – 3.5 8.5 – 3.5 > 9.5 Σ
Observed Frequency ni
Theoritical Frequency ei
6 34 132 179 218 183 146 69 33 1000
15.00 40.91 100.33 177.35 222.07 208.57 138.96 66.75 30.06 1000
(ni-ei)2
(ni-ei)2 / ei
81.00 47.75 1002.99 2.72 16.56 653.82 49.56 5.06 8.64
5.4 1.167 9.997 0.015 0.075 3.135 0.357 0.076 0.287 20.509
x2 distribution with f = 9-3 = 6 degree of freedom and α = 1% significance level, c0.99,6 = 16.812
(n i − ei ) 2 In this case ∑ = 20.509 > 16.812 ei i =1 9
So normal distribution is not a suitable one.
8.1 (a)
B = stress = y / 0.1;
A = strain = x / 10 ^
^
Let, E(B A = a) = α + β a Nos. 1 2 3 4 5 6 Σ
a=
^
β=
bi 10 20 30 40 50 60 210
ai 9x10-4 20x10-4 28x10-4 41x10-4 52x10-4 63x10-4 213x10-4
213 × 10 −4 = 35.5 × 10 − 4 , 6
b=
aibi 9x10-3 40x10-3 84x10-3 164x10-3 260x10-3 378x10-3 935x10-3
210 = 35 6
∑ a b − nab = 935 ×10 − 6 × 35 × 35.5 ×10 ∑ a − na 9619 ×10 − 6 × 35.5 ×10 −3
i
ai2 81x10-8 400x10-8 784x10-8 1681x10-8 2704x10-8 3969 x10-8 9619x10-8
−4
i
−8
2
2
2
−8
i
= 9.21 x 103 k/in2 ^
^
α = b − β a = 35 − 9.21× 10 3 × 35.5 ×10 − 4 = 35 − 32.7 = 2.3 ^
So Young’s modulus = β = 9.21 x 103 k/in2 ^
(b)
Assume E(B
A = a) = β a
6
∆2 = ∑ (bi − βa i ) 2 i =1
Now, 6 ∂∆ = ∑ 2(bi − β a i )(−a i ) = 0 ∂β i =1
or,
∑a b β= ∑a ^
i
i
2
i
935 × 10 −3 = = 9.72 × 10 3 ksi −8 9619 ×10
bi2 100 400 900 1600 2500 3600 9100
8.2 (a) Plot of stopping distance vs speed of travel
Stopping Distance, in m
50 45 40 35 30 25 20 15 10 5 0 0
20
40
60
80
100
120
140
Speed, in kph
(b)
Vehicle No.
Speed
Stopping Distance
(kph)
(m)
Xi
Yi
Xi2
Yi2
XiYi
Yi'=a+bXi
1
40
15
1600
225
600
13.43
2
9
2
81
4
18
1.46
3
100
40
10000
1600
4000
36.59
4
50
15
2500
225
750
17.29
5
15
4
225
16
60
3.78
6
65
25
4225
625
1625
23.08
7
25
5
625
25
125
7.64
8
60
25
3600
625
1500
21.15
9
95
30
9025
900
2850
34.66
10
65
24
4225
576
1560
23.08
11
30
8
900
64
240
9.57
(Y
12
Total:
125
45
15625
2025
5625
679
238
52631
6910
18953
46.25
1.552
71.716
On the basis of calculations in the above table we obtain the respective sample means of X and Y as,
X = 679/12 = 56.6 kph,
Y = 238/12 = 19.8 m
and corresponding sample variances,
1 (52631 − 12 × 56.6 2 ) = 1289.84 11 1 S y2 = (6910 − 12 × 19.8 2 ) = 200.50 11
S x2 =
,
From Eq. 8.4 & 8.3, we also obtain,
β=
18953 − 12 × 56.6 × 19.8 = 0.388 , 52631 − 12 × 56.6 2
α = 19.8 – 0.388×56.6 = -2.161
From Eq. 8.6a, the conditional variance is,
S
2 Y|X
1 n = ( y i − y i' ) 2 = 71.716 / (12 – 2) = 7.172 ∑ n − 2 i =1
and the corresponding conditional standard deviation is S Y | x = 2.678 m
From Eq. 8.9, the correlation coefficient is, n
1 ρˆ = n −1
∑x y i =1
i
i
− nx y
sx sy
=
1 18953 − 12 × 56.6 × 19.8 = 0.98 11 1289.84 ⋅ 200.50
(c) To determine the 90% confidence interval, let us use the following selected values of Xi = 9, 30, 60 and 125, and with t0.95,10 = 1.812 from Table A.3, we obtain, At Xi = 9;
< µY
X
> 0.90 = 1.33 ± 1.812 × 2.679
(9 − 56.6) 2 1 + = (−1.063 → 3.723)m 12 (52631 − 12 × 56.6 2 )
At Xi = 30;
< µY
X
> 0.90 = 9.48 ± 1.812 × 2.679
(30 − 56.6) 2 1 + = (7.708 → 11.252)m 12 (52631 − 12 × 56.6 2 )
At Xi = 60;
< µY
X
> 0.90 = 21.12 ± 1.812 × 2.679
(60 − 56.6) 2 1 + = (19.712 → 22.528)m 12 (52631 − 12 × 56.6 2 )
At Xi = 125;
< µY
X
(125 − 56.6) 2 1 > 0.90 = 46.34 ± 1.812 × 2.679 + = (43.220 → 49.460)m 12 (52631 − 12 × 56.6 2 )
8.3 (a) Plot of peak hour traffic vs daily traffic 1.6
1.4
Peak hour traffic (1000 vehicles)
1.2
1
0.8
0.6
0.4
0.2
0 0
1
2
3
4
5
6
7
Daily traffic (1000 vehicles)
(b)
Let X be the daily traffic volume and Y the peak hour traffic volume, both in thousand vehicles. By Eq. 8.9, the correlation coefficient between X and Y is estimated by n
1 ρˆ = n −1
∑x y i
i =1
i
− nx y
sx s y
n
With n = 5,
∑x y i =1
i
i
= 25.54, x = 1.02, y = 4.72, and the sample standard deviations sx
= 0.303315018 while sy = 1.251798706, ⇒ ρˆ = (1/4)(1.468 / 0.379689347) ≅ 0.967
(c) xi 5 4.5 6.5 4.6
yi 1.2 1 1.4 0.9
xiyi 6 4.5 9.1 4.14
xi2
y i' = αˆ + βˆxi
(yi – yi’)2
25 20.25 42.25 21.16
1.085577537 0.968474793 1.436885769 0.991895341
0.0130925 0.00099384 0.00136056 0.00844475
Sum Mean
3
0.6
1.8
9
23.6 4.72
5.1 1.02
25.54
117.66
0.61716656
0.00029469 0.02418634
Since we know the total (daily) traffic count (X), and want a probability estimate about the peak hour traffic (Y), we need the regression of Y on X, i.e. the best estimates of α and β in the model E(Y | X = x) = α + βx From Eq. 8.4 & 8.3 n
β=
∑x y i
i =1 n
∑x i =1
2 i
i
− nx y − nx
2
= (25.54 - 5×4.72×1.02) / (117.66 – 5×4.722) = 1.468 / 6.268 ≅ 0.234, and
α = y − βˆ x = 1.02 – 0.23420549×4.72 ≅ – 0.0854 Also, by (7.5), using the results in the table above, we estimate the variance of Y (assumed constant)
S Y2| x =
1 n ∑ ( yi − yi' ) 2 n − 2 i =1
= 0.02418634 / (5 – 2) ≅ 0.0081 Hence when the daily traffic is 6000 vehicles, i.e. x = 6, Y has the estimated parameters µY = –0.0854 + 0.234×6 = 1.319 σY = 0.0080621140.5 = 0.089789278 Hence the estimated probability of peak hour traffic exceeding 1500 vehicles is P(Y > 1.5) = 1 – P(Y ≤ 1.5) = 1 – Φ (1.5 − 1.319 ) = 1 – Φ(2.01) 0.09
≅
0.0222
8.4 (a)
Plot of per capita energy consumption vs per capita GNP 5000
Per Capita Energy Consumption
4500 4000 3500
Y
3000
Y'
2500
lower
2000
upper
1500 1000 500 0 0
2000
4000
6000
8000
10000
12000
Per Captia GNP
(b) Country # Per Capita GNP
Per Capita Energy Consumption
Xi
Yi
Xi2
Yi2
XiYi
Yi'=a+bXi
(Yi-Yi')2
1
600
1000
360000
1000000
600000
950.37
2463.269
2
2700
700
7290000
490000
1890000
1535.64
698286.73
3
2900
1400
8410000
1960000
4060000
1591.38
36624.478
4
4200
2000
17640000
4000000
8400000
1953.68
2145.238
5
3100
2500
9610000
6250000
7750000
1647.11
727412.94
6
5400
2700
29160000
7290000
14580000
2288.12
169643.90
7
8600
2500
73960000
6250000
21500000
3179.96
462341.12
8
10300
4000
106090000
16000000
41200000
Total:
37800
16800
252520000
43240000
99980000
3653.74
On the basis of calculations in the above table we obtain the respective sample means of X and Y as,
X = 37800/8 = 4725,
Y = 16800/8 = 2100
and corresponding sample variances,
1 (252520000 − 8 × 4725 2 ) = 10559285.71 , 7 1 S y2 = (43240000 − 8 × 2100 2 ) = 1137142.86 7
S x2 =
From Eq. 8.4 & 8.3, we also obtain,
β=
(c)
99980000 − 8 × 4725 × 2100 = 0.279 , α = 2100 – 0.279×4725 = 781.725 2 252520000 − 8 × 4725
From Eq. 8.9, the correlation coefficient is, n
1 ρˆ = n −1 (d)
∑x y i =1
i
i
− nx y
sx sy
=
1 99980000 − 8 × 4725 × 2100 = 0.85 7 10559285.71 ⋅ 1137142.86
From Eq. 8.6a, the conditional variance is,
S Y2| X =
1 n ( y i − y i' ) 2 = 2218817.515 / 6 = 369801.799 ∑ n − 2 i =1
and the corresponding conditional standard deviation is SY | X = 608.113
(e)
To determine the 95% confidence interval, let us use the following selected values of Xi = 600, 5400 and 10300, and with t0.975,6 = 2.447 from Table A.3, we obtain, At Xi = 600;
< µY
X
(600 − 4725) 2 1 > 0.95 = 950.37 ± 2.447 × 608.113 + = (62.263 → 1835.997) 8 (252520000 − 8 × 4725 2 )
At Xi = 5400;
< µY
X
> 0.95 = 2288.12 ± 2.447 × 608.113
(5400 − 4725) 2 1 + = (1749.407 → 2827.253) 8 (252520000 − 8 × 4725 2 )
At Xi = 10300;
< µY
(f)
X
> 0.95 = 3653.74 ± 2.447 × 608.113
(10300 − 4725) 2 1 + = (2556.391 → 4754.469) 8 (252520000 − 8 × 4725 2 )
Similarly, β = 2.588 , α = -709.673 and S X |Y = 1853.080 for predicting the per capita GNP on the basis of the per capita energy consumption.
8.5 (a)
Let X be the car weight in kips; X ~ N(3.33, 1.04). Hence X − µ 4.5 − 3.33 ) P(X > 4.5) = P( > σ 1.04 = P(Z > 1.125) = 1 – Φ(1.125) = 1 - 0.8697 ≅ 0.130
(b) Let Y be the gasoline mileage. For linear regression, we assume E(Y | X = x) = α + βx and seek the best (in a least squares sense) estimates of α and β in the model. Based on the following data,
sum average
xi (kips) 2.5 4.2 3.6 3.0
yi (mpg) 25 17 20 21
13.3 3.325
83.0 20.75
n
β=
∑x y i =1 n
i
∑x i =1
2 i
i
xiyi
xi2
y i' = αˆ + βˆxi
(yi – yi’)2
6.25 17.64 12.96 9.00
62.5 71.4 72 63
24.33641 16.94624 19.55453 22.16283
0.440358 0.002891 0.198442 1.352165
45.9
268.9
1.993856
− nx y − nx
2
= (45.9 - 4×3.325×20.75) / (268.9 – 4×3.3252) = 1.468 / 6.268 ≅ - 4.35, and
α = y − βˆ x = 20.75 – 4.347158218×3.325 ≅ 35.20 Also, by Eq. 8.6a, using the results in the preceding table, we estimate the variance of Y (which is assumed constant)
S
2 Y |x
1 n = ( y i − y i' ) 2 ∑ n − 2 i =1 = 1.993856 / (4 – 2) ≅ 0.997
Hence when the car weighs 2.3 kips, i.e. x = 2.3, Y has the estimated parameters µY = 35.20430108 + (-4.347158218)×2.3 = 25.206 σY = 0.9969278030.5 = 0.998
Hence the estimated probability of gas mileage being more than 28 mpg is P(Y > 28) = 1 – P(Y ≤ 28) = 1 – Φ ( 28 − 25.206 ) 0.998
= 1 – Φ(2.8) = 1 – 0.9974 ≅ 0.0026 To determine the 95% confidence interval, let us use the following selected values of Xi = 2.5 and 4.2, and with t0.975,2 = 4.303 from Table A.3, we obtain, At Xi = 2.5;
(c)
< µY
X
(2.5 − 3.3) 2 1 > 0.95 = 24.34 ± 4.303 × 0.998 + = (21.215 → 27.465)mpg 4 (45.85 − 4 × 3.3 2 )
At Xi = 30;
< µY
X
> 0.95 = 16.95 ± 4.303 × 0.998
(4.2 − 3.3) 2 1 + = (13.613 → 20.287)mpg 4 (45.85 − 4 × 3.3 2 )
Plot of gasoline mileage vs weight
30
Gasoline Mileage (mpg)
25
Y 20
Y' lower
15
upper 10 5 0 0
1
2
3 Weight (kip)
4
5
8.6 (a)
Let Y be the number of years of experience, and M be the measurement error in inches. We have the following data: i 1 2 3 4 5 Σ
mi 1.5 0.8 1 0.8 0.5 4.6
yi 3 5 10 20 25 63
yimi 4.5 4 10 16 12.5 47
yi2 9 25 100 400 625 1159
From Eq. 8.4 & 8.3, we have
β =
∑ y m − nym = -0.030010953, and ∑ y − ny i
i
2 i
2
α = m − βy = 1.298138007 so our regression model is Mmean = 1.298138007 - 0.030010953Y, Hence the mean measurement error when a surveyor has 15 years of experience is estimated to be 1.298138007 + (-0.030010953)(15) = 0.847973713, while the estimated variance of M is 0.073026652 (by equation 7.5), hence P(M < 1 | Y = 15) = Φ(
1 − 0.847973713 ) 0.073026652
= Φ(0.562571845) ≅ 0.713 (b)
No, our data for Y ranges from 3 to 25, so our regression model is only valid for predictions using Y values within this range. 60 is outside the range, where our model may not be correct. In fact, other factors could come into play, such as effects of aging—a 100-year-old surveyor might not be able to see the crosshair anymore!
8.7
(a)
Let E(DO
T=t) = α + βt T
DO
(days)
(ppm)
Xi
Yi
Xi2
Yi2
XiYi
Yi'=a+bXi
(Yi-Yi')2
1
0.5
0.28
0.25
0.0784
0.14
0.29
0.00012
2
1
0.29
1
0.0841
0.29
0.27
0.00052
3
1.6
0.29
2.56
0.0841
0.464
0.24
0.00260
4
1.8
0.18
3.24
0.0324
0.324
0.23
0.00246
5
2.6
0.17
6.76
0.0289
0.442
0.19
0.00048
6
3.2
0.18
10.24
0.0324
0.576
0.16
0.00027
7
3.8
0.1
14.44
0.01
0.38
0.14
0.00124
8
4.7
0.12
22.09
0.0144
0.564
0.09
0.00074
Total:
19.2
1.61
60.58
0.3647
3.18
0.00842
Σ ti = 19.2, Σ DOi = 1.61, Σ ti DOi = 3.180, Σ ti2 = 60.58, Σ DOi2 = 0.3647, ∆2 = Σ(DOi – DOi’)2 = 84.825 x 10-4
t= ^
19.2 = 2.4, 8
β=
DO =
Σt i ⋅ DOi − nt DO
^
Σt i − nt 2
2
=
1.61 = 0.20 8
3.18 − 8 × 2.4 × 0.20 = −0.0455 60.58 − 8 × 2.4 2
^
α = DO − β t = 0.20 + 0.0455 × 2.4 = 0.3092 So the least-squares regression equation is, ^
E (DO T=t) = 0.3092 – 0.0455 t
(b)
S x2 =
1 (60.58 − 8 × 2.4 2 ) = 2.0714 , 7
S y2 =
1 (0.3647 − 8 × 0.2 2 ) = 0.0058 7
n
1 ρˆ = n −1
S Y2| X =
∑x y i =1
i
i
− nx y
sx sy
=
1 3.18 − 8 × 2.4 × 0.2 = −0.89 7 2.0714 ⋅ 0.0058
1 n ∑ ( y i − y i' ) 2 = 0.00842 / 6 = 0.0014 n − 2 i =1
and the corresponding conditional standard deviation is SY | X = 0.037 ppm.
8.8
Cost vs floor area 250
thousand $
200 150 100 50 0 0
(a)
1
2 3 1000 sq ft
The standard deviation of Y is
4
0.0025 = 0.05. When X = 0.35, the mean of Y is
E(Y | X = 0.35) = 1.12×0.35 + 0.05 = 0.442, hence P(Y > 0.3 | X = 0.35) = 1 – P(Y ≤ 0.3 | X = 0.35) 0.3 − 0.442 ) = 0.9977 = 1 – Φ( 0.05 Also, as preparation for part (b), if X = 0.4 ⇒ E(Y | X = 0.4) = 1.12×0.4 + 0.05 = 0.498 ⇒ P(Y > 0.3 | X = 0.4) = 1 – P(Y ≤ 0.3 | X = 0.4) 0.3 − 0.498 ) = 1 – Φ( 0.05 = 0.999963 (b)
Theorem of total probability gives P(Y > 0.3) = P(Y > 0.3 | X = 0.35)P(X = 0.35) + P(Y > 0.3 | X = 0.4)P(X = 0.4) = 0.9977×(1/5) + 0.999963×(4/5) ≅ 0.9995
(c)
Let YA and YB be the respective actual strengths at A and B. Since these are both normal, YA ~ N(0.442, 0.05); YB ~ N(0.498, 0.05), Hence the difference D = YA –
YB ~ N(0.442 – 0.498,
D ~ N(–0.056, 0.005 ) Hence P(YA > YB) = P(YA – YB > 0) = P(D > 0)
0.05 2 + 0.05 2 ), i.e.
= 1 – Φ[ 0 − (−0.056) ] 0.005
= 1 – Φ(0.791959595) = 1 – 0.785807948 ≅ 0.214
8.9 (a)
The formulas to use are n
∑x y
β=
i
i =1 n
∑x i =1
2 i
− nx y
i
− nx
2
and
α = y − βˆ x The various quantities involved are calculated in the following table: n=
3 xi
yi
1.05 63 1.83 92 3.14 204 Sum = 6.02 359 Mean = 2.00666667 119.666667 Beta = Alpha =
154.676667 119.666667
xiyi
xi2
66.15 168.36 640.56 875.07
1.1025 3.3489 9.8596 14.311
/ -
2.23086667 139.131807
yi'
(yi-yi')2
53.3363865 93.3854267 107.417521 237.699949 198.246093 33.1074492 364.192825
= =
69.335 -19.465
The regression line y = –19.46514060 + 69.33478768 x can be plotted on the scattergram, and has the best fit to the data (in a least square sense). (b)
Note the words “for given floor area”. In this case, we should not simply calculate the sample standard deviation of the y data, since although the individual yi’s may deviate a lot from their mean, this may be due to changes in x and hence not really a randomness of y. Hence we need to calculate the conditional variance E(Y | x), assuming it is a constant. An unbiased estimate of this is given by Eq. 8.6a as
S Y2| x =
1 n ∑ ( yi − yi' ) 2 n − 2 i =1
= 364.192825 / (3 – 2) ≅ 364.193 Thus the standard deviation of construction cost for given floor area is estimated as SY|x = 364.1928250.5 ≅ 19.08 (c)
The individual sample standard deviations are sx = 1.056140773, sy = 74.46028024. Plugging these (and beta) into 8.10a,
ρˆ = βˆ
Sx Sy
= 69.3347877 ×1.056140773÷74.46028024 ≅ 0.983 Since it is close to 1, there is a strong linear relationship between X and Y (d)
When X = 2.5, µY = α + β(25) = –19.46514060 + 69.33478768(25) ≅ 153.8718286, thus Y ~ N(153.8718286, 19.08383675) ⇒ P(Y < 180 | X = 2.5) = P(Z < 180 − 153.871828 6 ) = Φ(1.369125704) ≅ 0.915 19.0838367 5
8.10 (a) Car #
Rated Actual Mileage Mileage (mpg)
(mpg)
Xi
Yi
Xi2
Yi2
XiYi
Yi'=a+bXi
(Yi-Yi')2
1
20
16
400
256
320
15.54
0.215
2
25
19
625
361
475
19.29
0.086
3
30
25
900
625
750
23.05
3.807
4
30
22
900
484
660
23.05
1.100
5
25
18
625
324
450
19.29
1.671
6
15
12
225
144
180
11.78
0.048
Total:
145
112
3675
2194
2835
6.927
On the basis of calculations in the above table we obtain the respective sample means of X and Y as,
X =145/6 = 24.2,
Y = 112/6 = 18.7
and corresponding sample variances,
S x2 =
1 (3675 − 6 × 24.2 2 ) = 34.17 , 5
1 S y2 = (2194 − 6 × 18.7 2 ) = 20.67 5
From Eq. 8.4 & 8.3, we also obtain,
β=
(b)
2835 − 6 × 24.2 × 18.7 = 0.751 , α = 18.7 – 0.751×24.2 = 0.512 2 3675 − 6 × 24.2
From Eq. 8.9, the correlation coefficient is, n
1 ρˆ = n −1
∑x y i =1
i
i
− nx y
sx sy
=
1 2835 − 6 × 24.2 ×18.7 = 0.97 5 34.17 ⋅ 20.67
From Eq. 8.6a, the conditional variance is,
S Y2| X =
1 n ( y i − y i' ) 2 = 6.927 / 4 = 1.732 ∑ n − 2 i =1
and the corresponding conditional standard deviation is SY | X = 1.316 mpg.
(c)
YQ = actual milage of model Q car YR = actual milage of model R car ∵ XQ = rated milage of model Q car = 22 mpg E(YQ) = 0.512 + 0.751x22 = 17.03 Similarly E(YR) = 0.512 + 0.751x24 = 18.54 Hence YQ = N(17.03, 1.32); YR = (18.54, 1.32) Assume YQ, YR to be statistically independent, P(YQ > YR) = P(YR - YQ < 0) = Φ
− (18.54 − 17.03)
1.32 2
= Φ (−0.809) = 0.21
(d)
To determine the 90% confidence interval, let us use the following selected values of Xi = 15, 20 and 30, and with t0.95,4 = 2.132 from Table A.3, we obtain, At Xi = 15;
< µY X > 0.90 = 11.78 ± 2.132 ×1.316
1 (15 − 24.2) 2 + = (9.446 → 14.114) mpg 6 (3675 − 6 × 24.2 2 )
At Xi =20;
< µY
X
(20 − 24.2) 2 1 > 0.90 = 15.54 ± 2.132 ×1.316 + = (14.066 → 17.014) mpg 6 (3675 − 6 × 24.2 2 )
At Xi = 30;
< µY
X
> 0.90 = 23.05 ± 2.132 ×1.316
(30 − 24.2) 2 1 + = (21.331 → 24.769) mpg 6 (3675 − 6 × 24.2 2 )
8.11 (a) Case No.
Obs. S'mt. Cal. S'mt. Yi
Xi
Yi2
Xi2
XiYi
Xi'=a+bYi
(Xi-Xi')2
1
0.7
4.8
0.49
23.04
3.36
-0.93
32.864
2
64
25.1
4096
630.01
1606.4
55.01
894.691
3
5
3.8
25
14.44
19
2.87
0.869
4
25
26.4
625
696.96
660
20.54
34.299
5
29.5
27.8
870.25
772.84
820.1
24.52
10.755
6
3.8
0.5
14.44
0.25
1.9
1.81
1.708
7
5.9
6.5
34.81
42.25
38.35
3.66
8.048
8
38.1
2.6
1451.61
6.76
99.06
32.12
871.498
9
185
174
34225
30276
32190
161.95
145.195
10
28.1
26.7
789.61
712.89
750.27
23.28
11.674
11
0.6
0.6
0.36
0.36
0.36
-1.02
2.628
12
29.2
31.5
852.64
992.25
919.8
24.26
52.484
13
32
31
1024
961
992
26.73
18.233
14
5.4
3.7
29.16
13.69
19.98
3.22
0.229
15
3.6
4
12.96
16
14.4
1.63
5.615
16
35.9
25.9
1288.81
670.81
929.81
30.18
18.291
17
11.6
11.3
134.56
127.69
131.08
8.70
6.757
18
12.7
14.2
161.29
201.64
180.34
9.67
20.495
19
46
42.1
2116
1772.41
1936.6
39.10
8.981
20
3.8
1.58
14.44
2.4964
6.004
1.81
0.052
21
4.9
5
24.01
25
24.5
2.78
4.932
22
11.7
11.8
136.89
139.24
138.06
8.79
9.066
23
1.83
3.39
3.3489
11.4921
6.2037
0.07
11.049
24
9.43
3.24
88.9249
10.4976
30.5532
6.78
12.552
25
6.6
4.3
43.56
18.49
28.38
4.28
0.000
Total:
600.4
491.8
48063.16
38138.51
41546.51
2182.967
On the basis of calculations in the above table we obtain the respective sample means of X and Y as,
Y = 600.4/25 = 24,
X = 491.8/25 = 19.7
and corresponding sample variances,
1 (48063.16 − 25 × 24 2 ) = 1401.91 24 1 S X2 = (38138.51 − 25 ×19.7 2 ) = 1185.98 24
S Y2 =
,
From Eq. 8.4 & 8.3, we also obtain,
β=
41546.51 − 25 × 24 × 19.7 = 0.884 , 48063.16 − 25 × 24 2
α = 19.7 – 0.884×24 = -1.551
Hence, E(X | y) = α + βy = -1.551 + 0.884y From Eq. 8.6a, the conditional variance is,
S X2 |Y =
1 n ( x i − x i' ) 2 = 2182.967 / (25 – 2) = 94.912 ∑ n − 2 i =1
and the corresponding conditional standard deviation is S X |Y = 9.742
(b)
From Eq. 8.9, the correlation coefficient is, n
1 ρˆ = n −1 (c)
∑x y i =1
i
i
− nx y
sx sy
=
1 41546.51 − 25 ×19.7 × 24 = 0.96 24 1185.98 ⋅ 1401.91
To determine the 95% confidence interval, let us use the following selected values of Yi = 64, 25, 5.9, 185, 0.6, 3.6, 11.6 and 46, and with t0.975,23 = 2.069 from Table A.3, we obtain, At Yi = 64;
< µ X Y > 0.95 = 55.01 ± 2.069 × 9.742
(64 − 24) 2 1 + = (49.048 → 60.975) 25 (48063.16 − 25 × 24 2 )
At Yi = 25;
< µ X Y > 0.95 = 20.54 ± 2.069 × 9.742 At Yi = 5.9;
(25 − 24) 2 1 + = (16.511 → 24.576) 25 (48063.16 − 25 × 24 2 )
(5.9 − 24) 2 1 + = (−0.833 → 8.159) 25 (48063.16 − 25 × 24 2 )
< µ X Y > 0.95 = 3.66 ± 2.069 × 9.742 At Yi = 185;
< µ X Y > 0.95 = 161.95 ± 2.069 × 9.742
(185 − 24) 2 1 + = (143.806 → 180.094) 25 (48063.16 − 25 × 24 2 )
At Yi = 0.6;
< µ X Y > 0.95 = −1.02 ± 2.069 × 9.742
(0.6 − 24) 2 1 + = (−5.804 → 3.761) 25 (48063.16 − 25 × 24 2 )
At Yi = 3.6;
< µX Y
(3.6 − 24) 2 1 > 0.95 = 1.63 ± 2.069 × 9.742 + = (−2.983 → 6.244) 25 (48063.16 − 25 × 24 2 )
At Yi = 11.6;
< µ X Y > 0.95 = 8.70 ± 2.069 × 9.742
(11.6 − 24) 2 1 + = (4.445 → 12.957) 25 (48063.16 − 25 × 24 2 )
At Yi = 46;
< µ X Y > 0.95 = 39.10 ± 2.069 × 9.742
(46 − 24) 2 1 + = (34.403 → 43.803) 25 (48063.16 − 25 × 24 2 )
Plot of calculated settlement vs observed settlement 200
Cal. S'mt.
150 Y Y' lower upper
100
50
0 0
50
100
-50
Obs. S'mt.
150
200
8.12 (a)
Plot of % loss in ridership vs % fare increase 16 % Loss in Ridership
14 12 Y
10
Y'
8
lower
6
upper
4 2 0 0
10
20
30
40
% Fare Increase
(b) Fare Loss in Increase Ridership %
%
Xi
Yi
Xi2
Yi2
XiYi
Yi'=a+bXi
(Yi-Yi')2
1
5
1.5
25
2.25
7.5
2.05
0.300
2
35
12
1225
144
420
11.55
0.202
3
20
7.5
400
56.25
150
6.80
0.491
4
15
6.3
225
39.69
94.5
5.22
1.176
5
4
1.2
16
1.44
4.8
1.73
0.282
6
6
1.7
36
2.89
10.2
2.36
0.442
7
18
7.2
324
51.84
129.6
6.17
1.070
8
23
8
529
64
184
7.75
0.063
9
38
11.1
1444
123.21
421.8
12.50
1.962
10
8
3.6
64
12.96
28.8
3.00
0.362
11
12
3.7
144
13.69
44.4
4.27
0.320
12
17
6.6
289
43.56
112.2
5.85
0.564
13
17
4.4
289
19.36
74.8
5.85
2.100
14
13
4.5
169
20.25
58.5
4.58
0.007
15
7
2.8
49
7.84
19.6
2.68
0.014
16
23
8
529
64
184
7.75
0.063
Total:
261.0
90.1
5757.00
667.23
1944.70
9.417
On the basis of calculations in the above table we obtain the respective sample means of X and Y as,
X =261/16 = 16.3,
Y = 90.1/16 = 5.6
and corresponding sample variances,
S x2 =
1 1 (5757 − 16 × 16.3 2 ) = 99.96 , S y2 = (667.23 − 16 × 5.6 2 ) = 10.66 15 15
From Eq. 8.4 & 8.3, we also obtain,
β=
(c)
1944.7 − 16 × 16.3 × 5.6 = 0.317 , α = 5.6 – 0.317×16.3 = 0.464 2 5757 − 16 ×16.3
From Eq. 8.9, the correlation coefficient is, n
1 ρˆ = n −1
∑x y i =1
i
i
− nx y
sx sy
=
1 1944.7 − 16 ×16.3 × 5.6 = 0.97 15 99.96 ⋅ 10.66
From Eq. 8.6a, the conditional variance is,
S Y2| X =
1 n ( y i − y i' ) 2 = 9.417 / 14 = 0.673 ∑ n − 2 i =1
and the corresponding conditional standard deviation is SY | X = 0.820
(d)
To determine the 90% confidence interval, let us use the following selected values of Xi = 4,
23, 38, 12 and 7, and with t0.95,14 = 1.761 from Table A.3, we obtain, At Xi = 4;
< µY
X
(4 − 16.3) 2 1 > 0.90 = 1.73 ± 1.761× 0.82 + = (1.147 → 2.315) 16 (5757 − 16 ×16.3 2 )
At Xi =23;
< µY
X
> 0.90 = 7.75 ± 1.761× 0.82
(23 − 16.3) 2 1 + = (7.311 → 8.188) 16 (5757 − 16 ×16.3 2 )
At Xi = 38;
< µY
X
(38 − 16.3) 2 1 > 0.90 = 12.5 ± 1.761× 0.82 + = (11.615 → 13.387) 16 (5757 − 16 ×16.3 2 )
At Xi = 12;
< µY
X
> 0.90 = 4.27 ± 1.761× 0.82
(12 − 16.3) 2 1 + = (3.870 → 4.661) 16 (5757 − 16 ×16.3 2 )
> 0.90 = 2.68 ± 1.761× 0.82
(7 − 16.3) 2 1 + = (2.181 → 3.183) 16 (5757 − 16 ×16.3 2 )
At Xi = 7;
< µY
X
8.13 (a) Given I = 6, E(D) = 10.5 + 15x6 = 100.5 Hence P(D>150) = 1 − Φ
(b)
(150 − 100.5) = 1 − Φ (1.65) = 1 − 0.95 = 0.05 30
Given I = 7, E(D) = 10.5 + 15x7 = 115.5 Given I = 8, E(D) = 10.5 + 15x8 = 130.5 Hence E(D) = E(D⏐6)x0.6 + P(D⏐7)x0.3 + E(D⏐7)x0.1 = 100.5x0.6 + 115.5x0.3 +130.5x0.1 = 108 million $
8.14 (a) Load
Deflection
tons
cm
Xi
Yi
Xi2
Yi2
XiYi
Yi'=a+bXi
(Yi-Yi')2
1
8.4
4.8
70.56
23.04
40.32
4.44
0.126
2
6.7
2.9
44.89
8.41
19.43
3.43
0.276
3
4.0
2.0
16
4
8
1.81
0.037
4
10.2
5.5
104.04
30.25
56.1
5.52
0.001
Total:
29.3
15.2
235.49
65.70
123.85
0.440
On the basis of calculations in the above table we obtain the respective sample means of X and Y as,
X =29.3/4 = 7.3,
Y = 15.2/4 = 3.8
and corresponding sample variances,
1 S x2 = (235.49 − 4 × 7.3 2 ) = 6.96 , 3
1 S y2 = (65.70 − 4 × 3.8 2 ) = 2.65 3
From Eq. 8.4 & 8.3, we also obtain,
β=
123.85 − 4 × 7.3 × 3.8 = 0.599 , 235.49 − 4 × 7.3 2
α = 3.8 – 0.599×7.3 = -0.591
From Eq. 8.6a, the conditional variance is,
S Y2| X =
1 n ( y i − y i' ) 2 = 0.44 / 2 = 0.220 ∑ n − 2 i =1
and the corresponding conditional standard deviation is SY | X = 0.469 cm.
(b)
To determine the 90% confidence interval, let us use the following selected values of Xi = 4 and 10.2, and with t0.95,2 = 2.92 from Table A.3, we obtain, At Xi = 4;
< µY
X
> 0.90 = 1.81 ± 2.92 × 0.469
(4 − 7.3) 2 1 + = (0.597 → 3.017) cm 4 (235.49 − 4 × 7.3 2 )
At Xi =10.2;
< µY
(c)
X
(10.2 − 7.3) 2 1 > 0.90 = 5.52 ± 2.92 × 0.469 + = (4.422 → 6.625) cm 4 (235.49 − 4 × 7.3 2 )
Under a load of 8 tons, E(Y) = -0.591 + 0.599x8 = 4.20 cm Let y75 be the 75-percentile reflection
P(Y < y75) = 0.75 = Φ
Hence
y 75 − 4.20 0.469
y 75 − 4.20 = Φ −1 (0.75) = 0.675 0.469
and
y 75 = 4.52cm
8.15 (a) Deformation
Brinell Hardness
mm
kg/mm2
Xi
Yi
Xi2
Yi2
XiYi
Yi'=a+bXi
(Yi-Yi')2
1
6
68
36
4624
408
68.65
0.419
2
11
65
121
4225
715
61.88
9.720
3
13
59
169
3481
767
59.18
0.031
4
22
44
484
1936
968
47.00
9.000
5
28
37
784
1369
1036
38.88
3.543
6
35
32
1225
1024
1120
29.41
6.699
Total:
115
305
2819
16659
5014
29.412
On the basis of calculations in the above table we obtain the respective sample means of X and Y as,
X =115/6 = 19.2,
Y = 305/6 = 50.8
and corresponding sample variances,
1 S x2 = (2819 − 6 × 19.2 2 ) = 122.97 , 5
S y2 =
1 (16659 − 6 × 50.8 2 ) = 230.97 5
From Eq. 8.9, the correlation coefficient is, n
1 ρˆ = n −1 (b)
∑x y i =1
i
i
− nx y
sx sy
From Eq. 8.4 & 8.3, we also obtain,
=
1 5014 − 6 ×19.2 × 50.8 = −0.99 5 122.97 ⋅ 230.97
β=
5014 − 6 ×19.2 × 50.8 = −1.353 , α = 50.8 + 1.353×19.2 = 76.765 2 2819 − 6 ×19.2
From Eq. 8.6a, the conditional variance is,
S Y2| X =
1 n ∑ ( y i − y i' ) 2 = 29.412 / 4 = 7.353 n − 2 i =1
and the corresponding conditional standard deviation is SY | X = 2.712 kg/mm2.
(c)
E(Y⏐X=20) = 76.765 – 1.353x20 = 49.7 Hence, P(40
Φ
≦
50)
50 − 49.7 40 − 49.7 −Φ = Φ (0.11) − Φ (−3.58) = 0.544 − 0.0002 = 0.544 2.71 2.71
=
8.16 (a) Travel Speed
Stopping Distance
mph
ft
Xi
Yi
Xi2
Yi2
XiYi
Yi'=a+bXi
(Yi-Yi')2
1
25
46
625
2116
1150
42.48
12.384
2
5
6
25
36
30
3.29
7.370
3
60
110
3600
12100
6600
111.07
1.152
4
30
46
900
2116
1380
52.28
39.437
5
10
16
100
256
160
13.08
8.502
6
45
75
2025
5625
3375
81.68
44.578
7
15
16
225
256
240
22.88
47.377
8
40
76
1600
5776
3040
71.88
16.993
9
45
90
2025
8100
4050
81.68
69.277
10
20
32
400
1024
640
32.68
0.465
Total:
295
513
11525
37405
20665
Car #
247.536
On the basis of calculations in the above table we obtain the respective sample means of X and Y as,
X =295/10 = 29.5,
Y = 513/10 = 51.3
and corresponding sample variances,
S x2 =
1 1 (11525 − 10 × 29.5 2 ) = 313.61 , S y2 = (37405 − 10 × 51.3 2 ) = 1232.01 9 9
From Eq. 8.9, the correlation coefficient is, n
1 ρˆ = n −1
∑x y i =1
i
i
− nx y
sx sy
=
1 20665 − 10 × 29.5 × 51.3 = 0.99 9 313.61 ⋅ 1232.01
We can say that there is a reasonable linear relationship between the stopping distance and the speed of travel. (b) Plot of stopping distance vs travel speed 120
Stopping Distance, in ft
100 80 60 40 20 0 0
10
20
30
40
Travel speed, in mph
(c)
E(Y X) = a + bx + cx2 ;
let x’ = x2
So E(Y X) = a + bx + cx’
Σx i = 295, Σx' i = 11525, Σy i = 492
Σ( x i − x i ) 2 = 2822.5, Σ( x' i − x') 2 = 1186.06 × 10 4 Σ( xi − x i )( x'i − x') = 177387.5, Σ( xi − x i )( yi − y ) = 5641 Σ( x'i − x')( yi − y ) = 35829.5
50
60
70
^
x = 29.5, x' = 1152.5, α = y = 49.2 6x2822.5 + cx177387.5 = 5641 and bx177387.5 + cx1186.06x104 = 358295
det b= det
det c=
5641
177387.5
358295 1186.06 × 10 4 2822.5
177387.5
=
3.34859 ×10 9 = 1.666 2.01022 × 10 9
177387.5 1186.06 × 10 4 2822.5
5641
177387.5 358295 2.01022 ×10 9
=
1.064475 ×10 7 = 0.0053 2.01022 ×10 9
^
α = 49.2 – 1.666x29.5 – 0.0053x1152.5 = 49.2 – 49.147 – 6.10825 = -6.055 So E(Y X) = -6.055 + 1.666x + 0.0053 x2 ∆2 = Σ(yi – yi’)2 = 376.551
s Y2 X = (d)
376.551 = 53.793 10 − 2 − 1
;
sY
X
= 7.334 ft
At a speed of 50 mph, the expected stopping distance is E(Y) = -6.055 + 1.66x50 + 0.053x502 = 90.2 Let y90 be the distance allowed P(Y < y90) = Φ (
y 90 − 90.2 ) ≡ 0 .9 7.334
Hence,
y 90 − 90.2 = Φ −1 (0.9) = 1.28 7.334
and
y 90 = 7.334 ×1.28 + 90.2 = 99.6 ft
8.17 (a) Population
Total Consumption 106 gal/day
Xi
Yi
Xi2
Yi2
XiYi
1
12,000
1.2
144000000
1.44
14400
-0.10
1.682
2
40,000
5.2
1600000000
27.04
208000
4.97
0.055
3
60,000
7.8
3600000000
60.84
468000
8.58
0.610
4
90,000
12.8
8100000000
163.84
1152000
14.00
1.452
5
120,000
18.5
14400000000
342.25
2220000
19.43
0.862
6
135,000
22.3
18225000000
497.29
3010500
22.14
0.025
7
180,000
31.5
32400000000
992.25
5670000
30.28
1.498
Total: 637000.0
99.3
78469000000.00 2084.95 12742900.00
Yi'=a+bXi (Yi-Yi')2
6.185
On the basis of calculations in the above table we obtain the respective sample means of X and Y as,
X = 637000/7 = 91000,
Y = 99.3/7 = 14.2
and corresponding sample variances,
1 (78469000000 − 7 × 91000 2 ) = 3417000000 , 6 1 S y2 = (2084.95 − 7 ×14.2 2 ) = 112.72 6
S x2 =
From Eq. 8.4 & 8.3, we also obtain,
β=
12742900 − 7 × 91000 ×14.2 = 0.000181 , α = 14.2 – 0.000181×91000 = 2 78469000000 − 7 × 91000
- 2.266
(b)
From Eq. 8.6a, the conditional variance is,
S
2 Y|X
1 n = ( y i − y i' ) 2 = 6.185 / 5 = 1.237 ∑ n − 2 i =1
and the corresponding conditional standard deviation is SY | X = 1.112
(c)
To determine the 98% confidence interval, let us use the following selected values of Xi = 12000, 90000 and 180000, and with t0.99,5 = 3.365 from Table A.3, we obtain (in the unit “x106 gal/day”), At Xi = 12000;
< µY
X
(12000 − 91000) 2 1 > 0.98 = −0.10 ± 3.365 ×1.112 + = (−2.600 → 2. 7 (78469000000 − 7 × 91000 2 )
At Xi =90000;
< µY
X
> 0.98 = 14 ± 3.365 ×1.112
(90000 − 91000) 2 1 + = (12.590 → 15.42 7 (78469000000 − 7 × 91000 2 )
At Xi = 180000;
< µY
(d)
X
> 0.98 = 30.28 ± 3.365 ×1.112
(180000 − 91000) 2 1 + = (27.554 → 32 7 (78469000000 − 7 × 91000 2 )
E(Y⏐X=100,000) = -2.266 + 0.00181x100000 = 15.8 P(Y > 17⏐X=100,000) = 1 – Φ (17 − 15.8 ) = 1 – Φ(1.08) = 0.14 1.112
8.18 E(X V,H) = αV
β1
H β2
Taking the logarithm on both sides of the above equations, ln E(X V,H) = ln α + β 1 ln V + β 2 ln H Let ln E(X V,H) = Y
ln α = β 0 , ln V = X 1 , ln H = X 2 Then Y = β0 + β1X1 + β2X2 To find β0, β1, and β2
Σx1i = 12.9, Σx 2i = 19.67, Σy i = −0.021 Σ( x1i − x i ) 2 = 4889 ×10 − 4 , Σ( x 2i − x 2 ) 2 = 25388.9 ×10 − 4 Σ( x1i − x i )( x 2i − x 2 ) = −558.5 × 10 −4 , Σ( x1i − x i )( y i − y ) = 5076.35 × 10 −4 Σ( x 2i − x 2 )( y i − y ) = −38680.89 × 10 −4
x 1 = 1.075, x 2 = 1.639, y = −0.00175 ^
^
β 1 × 4889.0 ×10 − 4 − β 2 558.5 × 10 − 4 = 5076.35 × 10 − 4 ^
^
− β 1 × 558.5 × 10 − 4 + β 2 25388.92 × 10 − 4 = −38680.89 × 10 − 4
^
β1 =
5076.35
− 558.5
− 38680.89
25388.92
4889
− 558.5
− 558.5
25388.92
=
1.0728 = 0.866 1.2381×10 8
^
β2 =
4889
5076.35
− 558.5
− 38680.89
1.2381×10 8
=
− 1.8628 ×10 8 = −1.50 1.2381×10 8
^
β 0 = -0.00175 – 0.866x1.075 + 1.5x1.639 = 1.53 Now ln α = B0 ^
So
α = e1.53 = 4.618
So
α = 4.618, β 1 = 0.866 and β 2 = -1.50
^
^
^
Mean Velocity
Mean Depth
Mean Oxygenation Rate
(fps)
(ft)
(ppm/day)
Vi
Hi
Xi
Xi'
(Xi-Xi')2
1
3.07
3.27
2.272
2.06
0.044
2
3.69
5.09
1.44
1.25
0.038
3
2.1
4.42
0.981
0.94
0.001
4
2.68
6.14
0.496
0.71
0.047
5
2.78
5.66
0.743
0.83
0.008
6
2.64
7.17
1.129
0.56
0.327
7
2.92
11.41
0.281
0.30
0.000
8
2.47
2.12
3.361
3.27
0.008
9
3.44
2.93
2.794
2.68
0.012
10
4.65
4.54
1.568
1.81
0.057
11
2.94
9.5
0.455
0.40
0.003
12
2.51
6.29
0.389
0.65
0.068
35.89
68.54
15.909
Total:
From Eq. 8.27, S X V , H =
∆2 0.612 = = 0.26 ppm / day n − k −1 12 − 2 − 1
0.612
8.19
(a)
Var (Y X ) = σ 2 x 4 ; ^
β=
Thus wi =
1 xi
4
Σwi (Σwi x i y i ) − (Σwi y i )(Σwi x i ) Σ wi ( Σ wi x i ) − ( Σ wi x i ) 2 2
40.18 × 10 −14 × 157.76 × 10 −7 − 76.03 × 10 −10 × 75.60 × 10 −11 = 40.18 ×10 −14 × 166.67 ×10 −8 − (75.60) 2 × 10 − 22 = 6.02 ^
Σw y − β Σwi x i 76.03 × 10 −10 − 6.02 × 75.60 × 10 −11 = α= i i Σ wi 40.18 × 10 −14 ^
= 7595.5 ^
^
To determine α and β ; wi (yi ^
2
yi wi wi xi wi yi wi xi yi wi xi xi 3 4 -14 -11 10 -7 (x10 ) (x10 ) (x10 ) (x10 ) (x10 ) (x10 ) (x10-8) 1.4 1.6 26.03 36.44 41.65 58.31 51.02 2.2 2.3 4.27 9.39 9.82 21.60 20.66 2.4 2.0 3.01 7.22 6.02 14.45 17.36 2.7 2.2 1.88 5.08 4.14 11.18 13.72 2.9 2.6 1.41 4.09 3.67 10.64 11.89 3.1 2.6 1.08 3.35 2.81 8.71 10.41 3.6 2.1 0.60 2.16 1.26 4.54 7.72 4.1 3.0 0.35 1.44 1.05 4.31 5.95 3.4 3.0 0.75 2.55 2.25 7.65 8.65 4.3 3.8 0.29 1.25 1.10 4.73 5.41 5.1 5.1 0.15 0.77 0.77 3.93 3.84 5.9 4.2 0.08 0.47 0.34 2.01 2.87 6.4 3.8 0.06 0.38 0.23 1.47 2.44 4.6 4.2 0.22 1.01 0.92 4.23 4.73 -14 -11 -10 -7 40.18x10 75.60x10 76.03x10 157.76x10 166.67x10-8 Σ
So,
E(Y X) = 7595.5 + 6.02 X
^
- α - β xi)2 (x10-6) 0 1.2599 0.1204 0.0609 0.0114 0.0010 0.4133 0.0185 0.0271 0.0587 0.2419 0.0010 0.0394 0.0988 2.3523x10-6
(b)
s 2Y sY
(c)
X
X
=
2.3523 × 10 −6 4 x = (0.1960 × 10 − 6 ) x 4 14 − 2
= (0.4427 ×10 −3 ) x 2
To determine the 98% confidence interval, let us use the following selected values of xi = 1.5, 2.5, 3.5 and 4.5, and with t0.99,12 = 2.681 from Table A.3, we obtain, At xi = 1.5;
< µY
X
> 0.98 = (7595.5 + 6.02 ×1.5 ×10 3 )
± 2.681× (0.4427 ×10 − 3 × (1.5 ×10 3 ) 2 )
(1.5 ×10 3 − 3721.4) 2 1 + 14 (220630000 − 14 × 3721.4 2 )
= (15274.53 → 17976.47) At xi = 2.5;
< µY
X
> 0.98 = (7595.5 + 6.02 × 2.5 × 10 3 )
± 2.681× (0.4427 ×10 −3 × (2.5 × 10 3 ) 2 )
(2.5 ×10 3 − 3721.4) 2 1 + 14 (220630000 − 14 × 3721.4 2 )
= (19999.82 → 25291.18) At xi = 3.5;
< µY
X
> 0.98 = (7595.5 + 6.02 × 3.5 ×10 3 )
(3.5 ×10 3 − 3721.4) 2 1 ± 2.681× (0.4427 ×10 × (3.5 ×10 ) ) + 14 (220630000 − 14 × 3721.4 2 ) −3
3 2
= (24730.18 → 32600.82) At xi = 4.5;
< µY
X
> 0.98 = (7595.5 + 6.02 × 4.5 × 10 3 )
± 2.681× (0.4427 ×10 −3 × (4.5 × 10 3 ) 2 ) = (27313.05 → 42057.95)
(4.5 ×10 3 − 3721.4) 2 1 + 14 (220630000 − 14 × 3721.4 2 )
(d)
If x = 3500 E(Y X) = 7595.5 + 6.02x3500 = 2.8665x10-4
sY
X
= 0.4427 × 10 −3 ⋅ (3500) 2 = 5423
Assuming Y is N(2.8665 x 104, 5423) P(Y > 30,000) = 1 - φ[(30000-28665) / 5423] = 1 - φ(0.25) = 0.40
9.1 (a) Let p = Probability that the structure survive the proof test. P(p=0.9) = 0.70, P(p=0.5) = 0.25, P(p=0.10) =0.05 P(survival of the structure) = P(S) = P(S⏐p=0.9) P(p=0.9) + P(S⏐p=0.5) P(p=0.5) + P(S⏐p=0.1) P(p=0.1) = 0.9×0.7 + 0.5×0.25 + 0.1×0.05 = 0.76 (b) Let E = One structure survive the proof test. P’’(p=0.9) = P(p=0.9⏐E) = P(E⏐p=0.9)×P(p=0.9) / P(E) = 0.9×0.7 / 0.76 = 0.83 Similarly, P’’(p=0.5) = 0.5×0.25 / 0.76 = 0.16 P’’(p=0.10) = 0.10×0.05 / 0.76 = 0.01 (c) P(p=0.9⏐E) =
∑p
i
P’’(p=pi) = 0.9×0.83 + 0.5×0.16 + 0.1×0.01
i
= 0.747 + 0.080 + 0.001 = 0.828 (d) Let A = event of two survival and one failure in three tests. So, P(pi⏐A) = P(A⏐pi) P(pi) / P(A) P(A) = P(A⏐p = 0.9) P(p=0.9) + P(A⏐p=0.5) P(p=0.5) + P(A⏐p=0.1) P(p=0.1)
3 2
3 2
3 2
= ( ) (0.9)2×0.1×0.7 + ( )(0.5)2×0.5×0.25 + ( )(0.1)2×0.9×0.05 = 0.1701 + 0.09375 + 0.00135 = 0.2652 So, P(P=0.9⏐A) = 0.1701 / 0.2652 = 0.64 P(P=0.5⏐A) = 0.09375 / 0.2652 = 0.35 P(P=0.1⏐A) = 0.00135 / 0.2652 = 0.01 E(P⏐A) = 0.9×0.64 + 0.5×0.35 + 0.1×0.01 = 0.576 + 0.175 +0.001 = 0.752
9.2 (a) Let E1 = The output will be acceptable in the first day. Applying Bayes’ Theorem, P(pi⏐ E1 ) = P(E1⏐pi)×P(pi) / P( E1 ) P(E1) = P(E1⏐p=0.4) P(p=0.4) + P(E1⏐p=0.6) P(p=0.6) + P(E1⏐p=0.8) P(p=0.8) + P(E1⏐p=1.0) P(p=1.0) = 0.6×0.25 + 0.4×0.25 + 0.2×0.25 + 0×1.0 = 0.30 So P’’(p=0.4) = P(p=0.4⏐E1) = 0.6×0.25 / 0.30 = 0.50 P’’(p=0.6) = 0.4×0.25 / 0.30 = 0.333 P’’(p=0.8) = 0.2×0.25 / 0.30 = 0.167 P’’(p=1.0) = 0 E(p⏐E1) = 0.4×0.5 + 0.6×0.333 + 0.8×0.167 = 0.533
(b) Let, E2 = one unacceptable out of three trials. P(E2) = P(E2⏐p=0.4) P(p=0.4) + P(E2⏐p=0.6) P(p=0.6) + P(E2⏐p=0.8) P(p=0.8) + P(E2⏐p=1.0) P(p=1.0)
3 2
3 2
3 2
3 2
= ( ) (0.4)2×0.6×0.25 + ( ) (0.6)2×0.4×0.25 + ( ) (0.8)2×0.2×0.25 + ( ) (1.0)2×0×0.25 = 3×0.25 (0.096 + 0.144 + 0.128) = 0.276 So P’’(p=0.4) = P(p=0.4⏐E2) = 3×0.096×0.25 / 0.276 = 0.260 P’’(p=0.6) = 3×0.144×0.25 / 0.276 = 0.392 P’’(p=0.8) = 3×0.128×0.25 / 0.276 = 0.348 P’’(p=1.0) = 0 E(p⏐E2) = 0.26×0.4 + 0.392×0.6 + 0.348×0.8 + 0 = 0.617
(c) Let E3 = Out of three trial, first two are satisfactory and the last one is unsatisfactory. P(E3) = P(E3⏐p=0.4) P(p=0.4) + P(E3⏐p=0.6) P(p=0.6) + P(E3⏐p=0.8) P(p=0.8) + P(E3⏐p=1.0) P(p=1.0) = (0.4)2×0.6×0.25 + (0.6)2×0.4×0.25 + (0.8)2×0.2×0.25 + 0 = 0.25 (0.096 + 0.144 + 0.128) = 0.092 So P’’(p=0.4) = P(p=0.4⏐E3) = 0.096×0.25 / 0.092 = 0.260 P’’(p=0.6) = 0.144×0.25 / 0.092 = 0.392 P’’(p=0.8) = 0.128×0.25 / 0.092 = 0.348 P’’(p=1.0) = 0 E(p⏐E3) = 0.26×0.4 + 0.392×0.6 + 0.348×0.8 + 0 = 0.617 The posterior distribution will be same as shown in part (b) since the likelihood function of the observed events E2 and E3 are the same.
9.3 P(HAHF) = P(HALF) = 0.3, P(LAHF) = P(LALF) = 0.2 (a) P( H A H F ) = P( H A H F ⏐ HAHF) P(HAHF) + P( H A H F ⏐ HALF) P(HALF) + P( H A H F ⏐ LAHF)
P(LAHF) + P( H A H F ⏐ LALF) P(LALF) = 0.3×0.3 + 0.4×0.3 + 0.2×0.2 + 0.25×0.2 = 0.3 (b) P(HAHF⏐ H A H F ) = P( H A H F ⏐HAHF) P(HAHF) / P( H A H F ) = 0.3×0.3 / 0.3 = 0.3
(c) P(HAHF⏐ L A L F ) = P( L A L F ⏐ HAHF) P(HALF) / P( L A L F ) Now using similar method as in (a), P( L A L F ) = 0.2×0.3 + 0.1×0.3 + 0.1×0.2 + 0.25×0.2 = 0.16 So,
P(HAHF⏐ L A L F ) = 0.2×0.3 / 0.16 = 0.375
Similarly, P(HALF⏐ L A L F F) = 0.10×0.3 / 0.16 = 0.1875 P(LAHF⏐ L A L F ) = 0.10×0.2 / 0.16 = 0.125 P(LALF⏐ L A L F ) = 0.25×0.2 / 0.16 = 0.3125
9.4 (a) P(X=2) = P(X=2⏐m=0.4) P(m=0.4) + P(X=2⏐m=0.8) P(m=0.8) = 0.4×0.5 + 0.8×0.5 = 0.6 P(m=0.4⏐ X=2) = 0.4×0.5 / 0.6 = 0.333 = P’’(m=0.4) P(m=0.8⏐ X=2) = 0.8×0.5 / 0.6 = 0.667 = P’’(m=0.8) (b) Let, N = Number of accurate measurements. P(N ≥ 2) = P(N ≥ 2⏐m=0.4) P(m=0.4) + P(N ≥ 2⏐m=0.8) P(m=0.8)
3 2
3 3
3 2
3 3
P(N ≥ 2⏐m=0.4) = P(N ≥ 2⏐3, 0.4) = ( )× (0.4)2×0.6 + ( )× (0.4)3×(0.6)0 = 0.352 Similarly, P(N ≥ 2⏐m=0.8) = P(N ≥ 2⏐3, 0.8) = ( )× (0.8)2×0.2 + ( )× (0.8)3×(0.2)0 = 0.896 Considering posterior distribution of m P(N ≥ 2) = 0.352×0.333 + 0.896×0.667 = 0.714
9.5 P( λ = 4) = 0.4, P( λ = 5) = 0.6 (a) Let E = Observed test results of bending capacity for the two tests. P(E) = P(E⏐ λ = 4) P( λ = 4) + P(E⏐ λ = 5) P( λ = 5) P(E⏐ λ = 4) = {
1 4.5 2 ) 4
4.5 − 2 ( e 42
⋅ dm} {
1 5.2 2 ) 4
5.2 − 2 ( e 42
⋅ dm}
= 0.02085 dm2 where dm = small increment of m around m.
and P(E⏐ λ = 5) = {
1 4.5 2 ) 5
4 .5 − 2 ( e 52
1 5.2
⋅ dm} {
5 .2 − 2 ( 5 ) 2 e ⋅ dm} 52
So, P(E) = (0.02085×0.4 + 0.01454×0.6) dm2 = 0.01706 dm2 So, P’’( λ = 4⏐E) = P’’( λ = 4) = 0.02085×0.4 / 0.01706 = 0.489 P’’( λ = 5) = 0.01454×0.6 / 0.01706 = 0.511 (b) f’’M(m) = fM(m⏐ λ = 4) P’’( λ = 4) + fM(m⏐ λ = 5) P’’( λ = 5) 1 m
1 m
=
m − 2 ( 5 )2 m − 2 ( 4 )2 e × 0 . 489 + e × 0.511 42 52
= 0.03056m ⋅ e
1 m − ( )2 2 4
2
(c) P(M<2) =
∫ 0.03056m ⋅ e
+ 0.02044m ⋅ e
1 m − ( )2 2 4
2
dm +
= 0.09675
1 m − ( )2 2 4
∫ 0.02044m ⋅ e
1 m − ( )2 2 5
0
0
= [-16×0.03056 ⋅ e
1 m − ( )2 2 5
2 0
] +[-25×0.02044 ⋅ e
1 m − ( )2 2 5
]
2 0
dm
9.6 (a) P( α =2) = P( α =3) = 0.5 Let F = errors found from two measurements. P(F) = P(F⏐ α =2) P( α =2) + P(F⏐ α =3) P( α =3)
2 1 2 2 (1- )de} { (1- )de} = 0 2 2 2 2 2 1 2 2 8 2 P(F⏐ α =3) = { (1- )de} { (1- )de} = de 3 3 3 3 81
P(F⏐ α =2) = {
Where de = small increment of e So, P(F) = 0 +
8 4 2 ×0.5 de2 = de 81 81
P’’( α =2⏐F) = P’’( α =2) = 0 and P’’( α =3) = 1 E( α ⏐F) = 2×0 + 3×1 = 3 cm (b) f’’( α ) = k • L( α ) • f’( α ) where f’’( α ) = 1.0 ; 2≤ α ≤3 L( α ) = {
2
α
(1-
1
)} {
α
2
2
(1-
α
)}
)(1 −
2
) ⋅1
α
So, f’’( α ) = k ⋅
4
α
2
(1 −
1
α
α
;
2≤ α ≤3
= 0, else where To find the constant k, 3
4
∫k α
2
(1 −
2
or, 4k [−
1
α
+
1
α
)(1 −
3 2α
2
−
2
α
) ⋅ dα = 1.0
]
3 = 1.0 2
1
)(1 −
2 3α
3
or, k = 14.71 So, f’’( α ) =
58.84
α
2
(1 −
α
= 0, elsewhere
2
α
) ;
2≤ α ≤3
E( α ⏐F) =
3
3
∫ αf ' ' (α )dα = ∫ 2
2
58.84
α
= 58.85 [ln(α ) +
(1 − 3
α
−
1
α
)(1 −
2
α
)dα
1 3 ] = 2.61 cm. α2 2
9.7 Since, where
f’’(ν ) = kL(ν ) f’(ν )
L(ν ) =
e
−
ν 12
(
ν
12 1!
)1
, f ' (ν ) =
0.271
ν
; 0.5 ≤ ν ≤ 20
= 0, elsewhere So, f’’(ν ) = k ⋅ e
−
ν 12
×
ν 12
×
0.271
ν
;
0.5 ≤ ν ≤ 20
= 0, elsewhere To determine the constant k, 20
ν
0.271 −12 k ⋅ ∫ 12 e dν = 1.0 ⎯⎯→ k = 4.79 0.5 So the posterior distribution of ν is, ν
ν
− 0.271 −12 f ' ' (ν ) = 4.79 × e = 0.1082e 12 ; 12
= 0, elsewhere
0.5 ≤ ν ≤ 20
9.8 (a) P(p > 0.9) = 0.7 and P(p > 0.9) = 0.3
(b) f’’(p) = k × L(p) × f’(p)
3 3
L(p) = ( ) p3 × (1-p)0 = p3 So, f’’(p) = k × p3 × 1/3; = k × p3 × 7; =0, elsewhere
12 10
0 ≤ p ≤ 0.9 0.9 ≤ p ≤ 1.0
f"(p)
8
4
To find k, 0.9
1
∫3
6
k × p 3dp +
0
1.0
2
0.9
0
3 ∫ k × 7 ⋅ p dp = 1.0
0
or, p 4 0.9 p 4 1.0 → k = 1.523 k[ ] + 7k [ ] = 1.0 ⎯⎯ 3 4 0 4 0.9
0.5
1
p
1
So, f’’(p) = 0.508 p3 ; = 10.663 p3 ; =0, elsewhere 0.9
(c) E ''( p ) =
0 ≤ p ≤ 0.9 0.9 ≤ p ≤ 1.0
1.0
∫ 0.508 p dp + ∫ 10.663 p dp 4
0
4
0.9
h
i
= 0.508[ = 0.933
p 5 0.9 p 5 1.0 ] + 10.663[ ] 5 0 5 0.9
9.9 P(ν =20) = 2/3, P(ν =15) = 1/3 (a) Let X = Number of occurrences of fire in the next year. P(X=20) = P(X=20⏐ν =15) P(ν =15) + P(X=20⏐ν =20) P(ν =20) =
e −15 (15) 20 1 e −20 (20) 20 2 × + × 20! 3 20! 3
= 0.0139 + 0.0592 = 0.0731
(b) P’’(ν =15) =
P( X = 20ν = 15) ⋅ P(ν = 15) P( X = 20)
=
0.0139 = 0.19 0.0731
Similarly, P’’(ν =20) =
0.0592 = 0.81 0.0731
(c) P(X=20) = P(X=20⏐ν =15) P’’(ν =15) + P(X=20⏐ν =20) P’’(ν =20) =
e −15 (15) 20 e −20 (20) 20 × 0.19 + × 0.81 20! 20!
= 0.0079 + 0.0720 = 0.0799
9.10
λ is the parameter to be updated with the prior distribution of: P '(λ = 1 5) = 1/ 3
P '(λ = 1 10) = 2 / 3
(a) Let ε =accidents were reported on days 2 and 5. The probability to observe ε is
P(ε ) = P(ε | λ = 1/ 5) P '(λ = 1/ 5) + P(ε | λ = 1/10) P '(λ = 1/10) 1 2 1 3 1 1 2 1 3 2 = × exp( − ) × × exp( − ) × + × exp( − ) × × exp( − ) × 5 5 5 5 3 10 10 10 10 3 −3 = 8.949 × 10 Combining the observation with the prior distribution of λ with Bayesian’s theorem, the posterior distribution of λ can be calculated as follows:
1 2 1 3 1 × exp( − ) × × exp( − ) × P(ε | λ = 1/ 5) P '(λ = 1/ 5) 5 5 5 5 3 = 0.548 = P "(λ = 1/ 5) = −3 P(ε ) 8.949 × 10
P "(λ = 1/10) = 1 − 0.548 = 0.452 (b) Let ε =no accidents will occur for at least 10 days after second accident.
P(ε ) = P(T > 10) = 1 − P(T ≤ 10) = 1 − P(T ≤ 10 | λ = 1/ 5) P "(λ = 1/ 5) − P(T ≤ 10 | λ = 1/10) P "(λ = 1/10) = 1 − (1 − e
−
10 5
) × 0.548 − (1 − e
−
10 10
) × 0.452
= 0.240 (c) The updated distribution of λ as prior distribution in this problem:
P '(λ = 1/ 5) = 0.548
P '(λ = 1/10) = 0.452
Let ε = after the second accident in 10th day no accident was observed until the 15th day. The probability to observe Z is:
P(ε ) = P(T > 5) = [1 − P(T ≤ 5 | λ = 1/ 5)]P '(λ = 1/ 5) + [ P (T ≤ 5 | λ = 1/10)]P '(λ = 1/10) −
5
= [1 − (1 − e 5 )] × 0.548 + [1 − (1 − e
−
5 10
)] × 0.452
= 0.202 + 0.274 = 0.476 Combining the observation with the prior distribution of λ with Bayesian’s theorem, the posterior distribution of λ can be calculated as follows:
P "(λ = 1 5) =
[1 − P(T ≤ 5 | λ = 1/ 5)]P '(λ = 1/ 5) 0.202 = = 0.424 P( Z ) 0.476
P "(λ = 1 10) = 1 − 0.424 = 0.575 (d)
1 1 1 1 P( n = nc ) = P ( n = nc | λ = ) P(λ = ) + P( n = nc | λ = ) P(λ = ) 5 5 10 10 nc nc 2 1 = × e −2 × 0.548 + × e −1 × 0.425 nc ! nc ! = 0.074 ×
1 P "(λ = ) = 5
2 nc 1nc + 0.156 × nc ! nc !
0.2nc nc ! P( n = nc )
0.449 ×
0.1nc 1 nc ! P "(λ = ) = 10 P( n = nc ) 0.385 ×
1 5
Set P "(λ = ) = P "(λ =
1 ) 10
We get nc = 1.076 . Since
nc is an integer, nc = 2 .
(e) Non-informative prior for λ is used here. Likelihood function is: L(λ ) = λ e −2 λ λ e −3λ = λ 2 e −5λ Posterior distribution:
f "(λ ) ∝ L(λ ) = λ 2 e −5λ As
∫
+∞
0
λ 2e −5λ dλ =
1 +∞ Γ(3) 2 = (5λ ) 2 e − (5λ )d(5λ ) = ∫ 25 0 25 25
Thus the posterior distribution is:
f "(λ ) = 12.5λ 2 e −5λ
9.11 M is the parameter to be updated. It’s convenient to prescribe a conjugate prior to the Poisson process. From the information given in the problem, the mean and variance of the gamma distribution of M is:
E '( µ ) =
k' k '/ v '2 = 10 δ '( µ ) = = 0.4 v' k '/ v ' ,
Thus, k ' = 6.25 , v ' = 0.625 (a) It follows that the posterior distribution of M is also gamma distributed. From the relationship given in Table 9.1 between the prior and posterior statistics, and the sample data, the posterior distribution parameter of M can be estimated as:
k " = k '+ 1 = 7.25 v " = v '+ 0.6 = 1.225 k " 7.25 E "( M ) = = = 5.918 v " 1.225
δ '( M ) =
k "/ v "2 = 0.371 k "/ v "
(b)
P( X = 0) = ∫
+∞
=∫
+∞
=∫
+∞
0
0
0
P( X = 0 | m) f '( m)dm (0.4m)0 −0.4 m v "( v " m) k " −1 − mv " ×e × × e dm 0! Γ(k ") e −0.4 m ×
= 0.209
1.225(1.225m )7.25−1 −1.225m dm ×e Γ(7.25)
9.12 Let X = compression index X is N( µ , 0.16) Sample mean x =
1 (0.75 + 0.89 + 0.91 + 0.81) = 0.84 4
(a) From Equation 8.13, f’’( µ ) = N ( x,
σ
)
n
= N(0.84, 0.16/2) = N(0.84, 0.08) (b) f’( µ ) = N( µ µ' , σ µ' ) = N(0.8, 0.2) L( µ ) = N(0.84, 0.08) = N ( x,
σ
)
n
From Equation 9.14, 9.15,
µµ = ''
x(σ µ' ) 2 + µ µ' σ 2 / n (σ µ' ) 2 + σ 2 / n
0.84 × (0.2) 2 + 0.8 × (0.08) 2 = 0.2 2 + 0.08 2
= 0.834 and,
σµ = ''
(σ µ' ) 2 (σ 2 / n) (σ µ' ) 2 + (σ 2 / n)
=
0.2 2 × 0.08 2 = 0.0743 0.2 2 + 0.08 2
So, f’’( µ ) = N(0.834, 0.0743) (c) P( µ < 0.95) = φ (
0.95 − 0.834 ) = φ (1.56) = 0.94 0.0743
9.13 (a) T is N( µ , 10) Sample mean = t = 65 min. n=5 So the posterior distribution of µ ,
f ' ' ( µ ) = N (t ,
σ
) = N (65,
10
n
) = N (65,4.47) min.
5
(b) We know, f’( µ ) = N(65,4.47) min. L( µ ) = N (60,
10
) = N(60, 3.16) min.
10
From Equation 8.14 and 8.15,
µ µ'' = T
σ µ'' = T
65 × 3.162 + 60 × 4.47 2 = 61.7 min . 4.47 2 + 3.162 4.47 2 × 3.16 2 = 2.58 min . 4.47 2 + 3.16 2
f’’( µ T) = N(61.7, 2.58) min.
(c) From Equation 9.17, fT(t) = N(61.7,
102 + 2.582 )
= N(61.7, 10.33) min. P(T > 80) = 1 − φ (
80 − 61.7 ) = 1 − φ (1.77) = 0.0384 10.33
9.14 (a) The sample mean θ =
sθ = 2
1 (32 o 04'+31o 59'+32 o 01'+32 o 05'+31o 57'+32 o 00' ) = 32 o 01' 6
1 {(3' ) 2 + (2' ) 2 + 0 + (4' ) 2 + (4' ) 2 + (1' ) 2 } = 9.2 6 −1
sθ = 3.03' = 0.05 o o
The actual value of the angle is N (32 01' ,
0.05 o
)
6 or, N(32.0167o, 0.0204o) The value of the angle would be 32.0167 ± 0.0204o or 32o01’ ± 1.224’ (b) The prior distribution of θ can be modeled as N(32o, 2’) or N(32o, 0.0333o). 9.14 and 9.15, the Bayesian estimate of the elevation is,
θ''=
32 × 0.0204 2 + 32.0167 × 0.0333 2 0.0204 2 + 0.0333 2
= 32.0121o = 32o0.726’ and the corresponding standard error is,
σ θ'' =
0.0204 2 ⋅ 0.0333 2 = 0.01740 o or 1.044' 2 2 0.0204 + 0.0333
Hence,1.044’ θ is 32o0.726’ ± 1.044’
Applying Eqs.
9.15 Assume, First measurement → L=N(2.15, σ) Second measurement → L=N(2.20, 2σ) and Third measurement → L=N(2.18, 3σ) Applying Eqs. 9.14 and 9.15 and considering First and Second measurements,
2.20 × σ 2 + 2.15 × (2σ ) 2 L' ' = = 2.16km σ 2 + (2σ ) 2 and,
σ 2 × (2σ ) 2 σ = = 0.8944σ σ 2 + (2σ ) 2 '' L
Now consider L’’ with the third measurement and apply Eq. 9.14 again,
L' ' ' =
2.16 × (3σ ) 2 + 2.18 × (0.8944σ ) 2 = 2.1616km (3σ ) 2 + (0.8944σ ) 2
9.16 Assume the prior distribution of p to be a Beta distribution, then,
E ' ( p) =
q' = 0.1 ----------------------------- (1) q'+ r '
and
Var ' ( p) =
a' r ' = 0.06 2 ---------(2) 2 (q'+ r ' ) (q '+ r '+1)
From Equations 1 and 2, we get, q’ = 2.4 and r’ = 9 × 2.4 = 21.6 From Table 9.1, for a binomial basic random variable, q’’ = q’ + x = 2.4 + 1 = 3.4 v’’ = v’ + n-x = 21.6 + 12 – 1 = 32.6 So E’’(p) = 3.4 / (3.4 + 32.6) = 0.0945
Var ' ' ( p) =
3.4 × 32.6 = 0.00231 (3.4 + 32.6) 2 (3.4 + 32.6 + 1)
9.17
µ is the parameter to be updated with the prior distribution parameter µµ' = 3
σ µ' = 0.2 × 3 = 0.6
(a) From the observation we get
t = N(
2 + 3 0.5 , ) = N (2.5,0.354) 2 2
From the relationship given in Table 9.1,
µµ'' =
0.354 2 × 3 + 0.62 × 2.5 = 2.629 0.62 + 0.354 2
σ µ'' =
0.3542 × 0.62 = 0.305 0.62 + 0.354 2
(b) From equation 9.17 the distribution of T is
T = N ( µµ" , σ 2 + σ µ'' ) = N (2.629, 0.62 + 0.3052 ) = N (2.629,0.673) 2
P(T < 1) = P (
T − 2.629 1 − 2.629 < ) = Φ ( −2.421) = 0.774% 0.673 0.673
9.18 (a) The parameter needs to be updated is p. It’s suitable to set a conjugate prior for the problem. Thus, suppose p is Beta distributed with parameter p’ and q’. With the information presented in the problem, we can get
E '( p ) =
q' q'r' = 0.5 Var '( p ) = = 0.05 2 q '+ r ' ( q ' + r ') ( q ' + r ' + 1) ,
Thus, q ' = r ' = 2 (b) Upon the observation the posterior distribution of p can be updated in terms of p and q as:
q " = q '+ 2 = 2 + 2 = 4 r " = r '+ n − x = 2 + 1 − 2 = 1 f "( p ) =
Γ(5) p3 = 4 p3 Γ(4) Γ(1)
(c) Let ε =win the next two bids 1
1
0
0
p (ε ) = ∫ P(ε | p ) f "( p )dp = ∫ p 2 i4 p 3dp = 2 3
9.19 Using conjugate distributions, we assume Gamma distribution as prior distribution of λ. From Table 9.1, for an exponential basic random variable,
E ' (λ ) =
k' = 0.5; v'
Var ' (λ ) =
k ' 0 .5 = = (0.5 × 0.2) 2 v' v' 2
So, v’ = 50 and, k’ = 25 Now, v’’ = v’ +
∑x
i
= 50 + (1 + 1.5) = 52.5
k’’ = k’ + n = 25 + 2 =27 E’’(λ) = k’’/v’’ = 27/52.5 = 0.514 Var’’(λ) = k’’/v’’2 = 27/52.52 = 9.796 x 10-3 σ’’(λ) = 0.099
δ λ'' =
0.099 = 0.193 0.514
9.20
λ is the parameter to be updated. Let X denote the crack length. The probability of crack length larger than 4 is:
P( X > 4) = 1 − P ( X < 4) = e −4 λ The probability of crack length smaller than 6 is:
P( X < 6) = 1 − e −6 λ Likelihood function is
L(λ ) = λ e −3λ iλ e −5λ i(1 − e −6 λ )iλ e −4λ i e −4 λ iλ e −8λ = λ 4 e −24 λ (1 − e −6λ ) non-informative prior is used, thus
f "(λ ) ∝ L(λ ) = λ 4 e −24 λ (1 − e −6 λ )
∫
+∞
0
λ 4e −24λ (1 − e −6λ )dλ = 1/ 23415
thus f "(λ ) = 23415λ 4 e −24 λ (1 − e −6 λ )
9.21 (a) The occurrence rate of the tornado is estimated from the historical record as 0.1/year. The probability of x occurrences during time t (in years) is
P( X = x ) =
(0.1t ) x −0.1t e x!
The probability that the tornado hits the town during the next 5 years is
P( X > 0) = 1 − P( X = 0) = 1 − e −0.5 = 0.393 The prior distribution parameter of mean maximum wind speed is:
µµ' =
145 + 175 = 160 2
µµ' − 145 µµ' − 145 15 ' , thus σ = = = 7.653 = 1.96 µ 1.96 1.96 σ µ' The updated the distribution parameter of mean maximum wind speed is
µµ" =
160 ×
152 + 175 × 7.6532 2 = 164.854 152 2 + 7.653 2
152 × 7.6532 " 2 = 6.143 σµ = 152 2 + 7.653 2 From equation 9.17, the updated the distribution of maximum wind speed is
v = N ( µµ" , σ µ" + σ 2 ) = N (164.854, 152 + 6.1432 ) = N (164.854,16.209) 2
P( v > 220) = 1 − P( v ≤ 220) = 1 − P(
T − 164.854 220 − 164.854 < ) = 1 − Φ (3.402) = 0.033% 16.209 16.209
(3) The probability that the house will be damaged in the next 5 years is
P( v > 120, x > 0) = P ( v > 220) P( x > 0) = 0.876 × 0.393 = 0.344 where
P( v > 120) = 1 − P (v ≤ 120) = 1 − P(
T − 161.376 120 − 161.376 < ) = 1 − Φ ( −1.157) = 0.876 35.756 35.756
P( x > 0) = 1 − P( x = 0) = 1 − e −0.1×5 = 0.393 The probability that the house will not be damaged is 1 − 0.344 = 0.656
9.22 h is the parameter to be updated. (a) 10
10
0
0
E ( x ) = ∫ E ( x | h ) f ( h )dh = ∫ 0.5hi0.003h 2 = 3.75 where h
h
0
0
E ( x | h ) = ∫ xf ( x | h )dx = ∫
x dh = 0.5h h
(b) (i) the range of h is 4
1 h
L(h | x = 4) = p( x = 4 | h ) = Prior:
f '( h ) = 0.003h 2 Thus the posterior is
1 f "( h ) ∝ 0.003h 2 i = 0.003h h 10 1 Since ∫ 0.003h 2 i dh = 0.132 , the posterior is: 4 h 1 f "( h ) = i0.003h = 0.0238h 0.132 (iii) The probability that the hazardous zone will not exceed 4km in the next accident is: 4
P( X < 4) = ∫ f ( x )dx = 0.1428 × 4 = 0.5712 0
Where 10
10
4
4
f ( x ) = ∫ f ( x | h ) f "( h )dh = ∫
1 i0.0238hdh = 0.1428 h
9.23 (a) Let x denote the fraction of grouted length. Assumptions: the grouted length is normally distributed; the mean grouted length is normally distributed; the variance of fraction of length grouted to the constant and can be approximated by the observation data. From the information in the problem, we have
0.75 + 0.95 µµ = = 0.85 2 , '
Thus, σ µ = '
µµ' − 0.75 1.96
=
µµ' − 0.8 = 1.96 σ µ'
0.85 − 0.75 = 0.05 1.96
σ ≈ sL = 0.06
µµ" =
σ µ" =
0.85 ×
0.062 + 0.89 × 0.052 4 ≈ 0.88 0.062 2 + 0.05 4
0.062 × 0.052 4 = 0.026 0.062 2 + 0.05 4
The posterior covariance is 0.026/0.88=3%
X = N ( µµ" , σ 2 + σ "2µ ) = N (0.88,0.0654) (b) The probability the sample average grouted length is lessen than 0.85 is
P( X < 0.85) = Φ (
0.85 − 0.88 ) = 0.258 0.0654 2
The probability that the minimum grouted length is smaller than 0.83 is
1 − P( X 1 > 0.83) P( X 2 > 0.83) = 1 − 0.7782 = 0.395 So the second requirement is more stringent.
9.24 (1) Prior Non-informative prior is used as (P.M. Lee. Bayesian Statistics: An Introduction. Edward Arnold, 1989)
f '( µT , σ T2 ) ∝
1
σ T2
(2) Likelihood
, tn | µT , σ T )
p(t1 , t2 ,
⎡ (t − µ ) 2 ⎤ ⎡ (t − µ ) 2 ⎤ ⎡ (t − µ ) 2 ⎤ 1 1 exp ⎢ − 1 2T ⎥i exp ⎢ − 2 2T ⎥i i exp ⎢ − n 2T ⎥ 2σ T ⎦ 2πσ T 2σ T ⎦ 2σ T ⎦ 2πσ T 2πσ T ⎣ ⎣ ⎣ 1
= =
=
(
⎡ ∑ (ti − µT ) 2 ⎤ exp ⎢ − ⎥ n 2σ T2 2πσ T ⎣ ⎦
(
⎡ ( n − 1) s 2 + n ( t − µ ) 2 ⎤ exp n ⎢− ⎥ 2σ T2 ⎣ ⎦ 2πσ T
1
1
) )
where s 2 =
1 n (ti − t ) 2 ∑ n − 1 i =1
(3) Joint posterior is:
⎡ ( n − 1) s 2 + n ( t − µ ) 2 ⎤ f "( µT , σ T2 ) ∝ σ T− n −2 exp ⎢ − ⎥ 2σ T2 ⎣ ⎦ (4) Marginal posterior of σ T2
f "(σ T2 ) = ∫ f "( µT , σ T2 )d µT ⎡ ( n − 1) s 2 + n ( y − µ ) 2 ⎤ ∝ ∫ σ T− n −2 exp ⎢ − ⎥ d µT 2σ T2 ⎣ ⎦ ∝ (σ
)
2 − n −1 2 T
⎡ ( n − 1) s 2 ⎤ exp ⎢ − 2σ T2 ⎥⎦ ⎣
Which is a scaled inverse- χ 2 density
( )
So σ T2
"
= Inv − χ 2 ( n − 1, s 2 )
(5) Marginal posterior of µT
Similarly, the posterior of µT can be written as
f "( µT ) = ∫ f "( µT , σ T2 )dσ T2 ⎡ ( n − 1) s 2 + n ( t − µ ) 2 ⎤ 2 ∝ ∫ σ T− n −2 exp ⎢ − ⎥ dσ T 2σ T2 ⎣ ⎦ ∞
∝ A−n 2 ∫ z ( n −2) 2 exp( − z )dz 0
∝ ⎡⎣( n − 1) s 2 + n ( µT − t ) 2 ⎤⎦ ⎡ n( µT − t ) 2 ⎤ ∝ ⎢1 + ( n − 1) s 2 ⎥⎦ ⎣
−n / 2
−n / 2
which is the tn −1 ( t , s 2 n ) density. Or it can be written as
µT" − t s
n
= tn −1
Where z =
A 2σ
2 T
, A = ( n − 1) s 2 + n( µT − t ) 2
(6) We need both sample mean and sample standard deviation in this problem. From the information available in the problem, only the first set of data with 5 flight time has both sample mean and sample standard deviation. So only this set of data will be used.
s = 10 , n = 5 , t = 65 , S = (n − 1) s 2 = (5 − 1) × 102 = 400
µT" = tn −1 ( t , s 2 n ) = t4 (65, 4.472) E ( µT" ) = 65 , Var ( µT" ) =
(σ )
2 " T
4 × 4.4722 = 39.998 4−2
= Inv − χ 2 ( n − 1, s 2 ) = Inv − χ 2 (4,100)
E ⎡⎢(σ T2 ) ⎣
"
⎤ = 4 × 100 = 200 ⎥⎦ 4 − 2
" 2 × 42 Var ⎡⎢(σ T2 ) ⎤⎥ = × 100 = ∞ ⎣ ⎦ (4 − 2) 2 (4 − 4)
9.25 Let x denotes the rated value, and y denotes the actual value. Based on 9.24 to 9.27, α = 0.512 , β = 0.751 , σ 2 = 1.732 , s x2 = 34.167 From equation 9.26, E (Y | x ) = 0.512 + 0.751x The variance can be calculated by equation 9.34 as
6 −1 ⎧ 1 ⎡ 6 ( x − 24.167) 2 ⎤ ⎫ 2 Var (Y | x ) = × ⎨1 + ⎢1 + ⎬ × 1.732 = 3.368 + 0.0169( x − 24.167) ⎥ 6 − 3 ⎩ 6 ⎣ 6 − 1 34.167 ⎦ ⎭ E (Y | x = 24) = 0.512 + 0.751 × 24 = 18.536 Var (Y | x = 4) = 3.368 + 0.0169(24 − 24.167) 2 = 3.368 P(Y < 18 | x = 24) = Φ (
18 − 18.536 ) = 0.385 3.368
if the value of the basic scatter σ 2 is equal to 1.73, the variance of Y at x = 24 becomes 1.73. Then
P(Y < 18 | x = 24) = Φ (
18 − 18.536 ) = 0.34 1.73
10.1 (a) Apply Equation 10.2 g(0.10) ≤ 0.05 For n = 20, and r = 1
⎛ 20 ⎞ ⎛ 20 ⎞ ⎟⎟(0.10) 0 (0.90) 20 + ⎜⎜ ⎟⎟(0.10) 1 (0.90) 19 ⎝1⎠ ⎝0⎠
g(0.10) = ⎜⎜
= 0.1216 + 0.2702 = 0.05 The welds should not be accepted. r
(b) g(0.10) =
⎛ n⎞
∑ ⎜⎜ x ⎟⎟(0.03) x =0
x
(0.97) n − x = 0.90
x
(0.90) n − x = 0.05
⎝ ⎠
and from part (a) r
g(0.10) =
⎛n⎞
∑ ⎜⎜ x ⎟⎟(0.10) x =0
⎝ ⎠
From trial and error, and using Table of Cumulative Binomial Probabilities, n = 105, r = 5. (c) AOQ = pg(p) Here, n = 25 and r = 1 So, 1
AOQ = p
⎛ 25 ⎞
∑⎜ x ⎟ p x =0
⎝
⎠
x
(1 − p ) 25− x = p ⋅ (1 − p ) 25 + p 2 × 25 × (1 − p ) 24
This is plotted in Fig 10.1. From the graph, AOQL = 0.0332
Fig. 10.1 AOQ Curve
10.2 P(SO2 < 0.1 unit) ≥ 0.90 with 95% confidence i.e. to achieve a reliability of 0.9 with 95% confidence
n=
ln(0.05) = 28.4 ≈ 29days ln(0.90)
(a) Rn = 1 – C Here, n = 10 and R = 1 - 0.15 = 0.85 So, C = 1 - Rn = 1 - (0.85)10 = 0.8031 (b) C = 0.99, R = 0.85, n =
ln(1 − C ) ln(0.01) = = 28.3 ≈ 29 ln R ln(0.85)
So 19 additional borings should be made.
10.4
n=
ln(1 − C ) ln R
Here, C = 0.95 and R = 0.99 So,
n=
ln(1 − 0.95) = 298 ln(0.99)
298 consecutive successful starts would be required to meet the standard.
⎛ n⎞
r
g(0.05) =
∑ ⎜⎜ x ⎟⎟(0.05) x =0
x
⎝ ⎠
(0.95) n − x = 0.94
and, r
g(0.20) =
⎛n⎞
∑ ⎜⎜ x ⎟⎟(0.20) x =0
⎝ ⎠
x
(0.80) n − x = 0.10
The values of n and r are to be determined by trial and error. Alternatively, from fig. 10.1e, n = 30 and r = 3 approximately satisfy the above two equations.
10.6 (a) p = fraction defective = 0.30 n = 5, r = 0 From Equation 10.2,
⎛ 5⎞
g(0.3) = ⎜⎜ ⎟⎟(0.3) 0 (0.7) 5 = (0.7) 5 = 0.168 0
⎝ ⎠
= consumer risk
⎛ 5⎞
(b) g(0.1) = ⎜⎜ ⎟⎟(0.1) 0 (0.9) 5 = 0.59 0
⎝ ⎠
∴ P(rejection) = 1 – g(0.1) = 0.41
(a) α = Producer’s risks = P ( X < L µ a ) = Φ (
L − µa
σ
) = Φ(
118 − 120 ) = Φ (−1.58) = 0.057 4
n 10 L − µt 118 − 114 β = Consumer’s risks = P ( X > L µ t ) = 1 − Φ ( ) = 1 − Φ( ) = 1 − Φ (3.16) = 0.000792 σ 4 n (b) α = 0.10 = Φ (
L − 120 ) 4
n L − 114 β = 0.05 = 1 − Φ ( ) 4
(
or
or
10
L − 120 ) n = −1.28 4 n(
L − 114 ) = 1.64 4
n ∴
L − 120 − 1.28 = → L = 117.37 L − 114 1.64
From this, n= 3.8, so take n = 4.0 and L = 117.34 lb/ft3 (c) Using equations 10.10 and 10.11,
n ( L − 120) = −t 0.10, n −1 4
and
n ( L − 114) = t 0.05, n −1 4
After a trial-and-error procedure, n is found to be 6 and the corresponding value of L becomes 117.59 lb/ft3 or 117.29 lb/ft3. So take the average L in order to be fair to both consumer and producer and L = 117.44 lb/ft3.
10.8 α = 0.10, β = 0.10 Pa = 0.02, Pt = 0.10 The required sample size n is, 2
⎡ Φ −1 (1 − β ) − Φ −1 (α ) ⎤ ⎡ Φ −1 (0.90) − Φ −1 (0.10) ⎤ = n = ⎢ −1 ⎥ ⎢ −1 ⎥ −1 −1 ⎣ Φ (0.10) − Φ (0.02) ⎦ ⎣ Φ ( Pt ) − Φ ( Pa ) ⎦ 2
2
2
⎡ 1.28 − (−1.28) ⎤ ⎡ 2.56 ⎤ =⎢ =⎢ ⎥ ⎥ = 10.77 ≈ 11 ⎣ 0.78 ⎦ ⎣ − 1.28 − (−2.06) ⎦ The corresponding tolerable fraction defective M is
⎡ −1 ⎡ −1 Φ −1 (α ) ⎤ Φ −1 (0.10) ⎤ M = Φ ⎢Φ ( Pa ) − ⎥ = Φ ⎢Φ (0.02) − ⎥ n ⎦ 11 ⎦ ⎣ ⎣ ⎡ 1.28 ⎤ = Φ ⎢− 2.06 + ⎥ = Φ (−1.674) = 0.047 11 ⎦ ⎣
μa = 0.90, μt = 0.80, α = 0.05, β = 0.05, σ = 0.04
Φ(
L − 0.90 ) = 0.05 0.04
n(
or
L − 0.90 ) = −1.64 0.04
n and 1 − Φ (
L − 0.80 ) = 0.05 0.04
or
n(
L − 0.80 ) = 1.64 0.04
n Solving the above two equations, n = 2 and L = 0.85.